{"authorId":"","cover":"/content/images/2026/10/9e08d4941150-social-background.png","heroes":{"59ab1e54b3a5386e6900ca8678d4e839":"/content/images/2026/10/b1e12503f058-nullg-rl-full-preview.png","71c64fb16098ef3fc7196432defc52c7":"/content/images/2026/10/8096ea5e04f8-nullg-soundtrack-zero-g-chokehold.png","729c9a4edc92eaa4171ac274dda70ca2":"/content/images/2019/11/IMG_9531.JPG","7a23b4fc9f63f7327ef52913289d4327":"/content/images/2017/12/20171201-4K4X9516.jpg","7b367212a739706ba4e3e9b999bec15d":"/content/images/2026/10/5febc720f2c6-nullg-rl-hero.png","901f2fab2646e83671fe8f6e5500c117":"/content/images/2018/11/IMG_6734.jpg","94b2de371382ec793a64ed49343c14c6":"/content/images/2017/11/IMG_3194.JPG","9f1711a950f76598b42e3b1fd15b45f0":"/content/images/2026/10/44526a0d5fab-nullg-blender-fleet.png","b00b035b3794d4eb5846a7122bf026cf":"/content/images/2026/10/ab60de3338bd-nullg-weekly-fresh-xl-shield.png"},"kind":"index","logo":"/content/images/2016/08/projectable-me-4.png","nav":[{"id":"62e854d471683dea63488bcbca5c6188","label":"Home","url":"/","sortOrder":0},{"id":"1bc84fb15620acd635df7735a0b8ddd8","label":"Projects","url":"/projects/","sortOrder":1},{"id":"3207e2fec51d66d2a5b3d8d04feb9dd8","label":"Sim Racing","url":"/tag/sim-racing/","sortOrder":2},{"id":"59115afd52e59a74faa4b386e00ae3fa","label":"About","url":"/about-me/","sortOrder":3}],"page":1,"perPage":10,"posts":[{"id":"b00b035b3794d4eb5846a7122bf026cf","siteId":"projectable","slug":"nullg-weekly-progress-xl-fleet-rl-pulse-pure","title":"Shields up: new ships, Pulse on the grid, and a look inside NullG","markdown":"[NullG](https://projectable.me/projects/nullg/) had a busy week, and the best way to catch up is to see it moving. Five rebuilt XL ships, a fresh batch of learned pilots, shields rippling around the hulls, and a playable slice of WipEout Pulse: there's a lot to get into.\n\nThe new **Shields Up** video pulls the XL work together over the full **Warehouse Rage** track. Every ship gets screen time, every weapon gets a turn, and the camera gets close enough to enjoy the details. Turn the sound up for this one.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/2VYHD8xlsoM\" title=\"NullG — Shields Up: weekly development showcase\" loading=\"lazy\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" allowfullscreen\u003e\u003c/iframe\u003e\n\n[Watch Shields Up on YouTube](https://www.youtube.com/watch?v=2VYHD8xlsoM). It's 3 minutes and 19 seconds of this week's work, with a separate 90-second portrait edit for a quicker look.\n\n## The fleet is coming together\n\n**FEISAR, AG Systems, Auricom, Qirex and Piranha** all feature in the updated XL fleet. The familiar shapes are there, now with rebuilt geometry, physically based materials, and more attention to the cockpits, engines and moving control surfaces.\n\n![FEISAR and AG Systems in fresh native model-viewer captures.](https://projectable.me/content/images/2026/10/0f9bf03cf410-nullg-weekly-fresh-fleet-a.png)\n\n*FEISAR and AG Systems in the native Model Viewer.*\n\nThere's a lovely amount of character in these ships. Qirex's broad shape and Piranha's narrow profile ask very different things of the model work. Even the ailerons need care: their position, thickness and range of movement all affect how convincing the ship looks when it banks into a corner.\n\nThis week brought more of those details together, including clearance checks on Qirex and refinements to the thin control surfaces on AG Systems and Piranha. Lighting and materials carry the work from the menus onto the track, through bright open sections, tunnels and city lights.\n\n![Auricom, Qirex and Piranha in fresh native model-viewer captures.](https://projectable.me/content/images/2026/10/a300c3095728-nullg-weekly-fresh-fleet-b.png)\n\n*Auricom, Qirex and Piranha, with control surfaces posed in the engine viewer.*\n\nThe [Blender fleet article](https://projectable.me/nullg-recreating-wipeout-xl-models-in-blender/) gets into the reconstruction process. The video is the payoff: all five ships, moving through the world they were built for.\n\n## Fresh pilots, straight from training\n\nThe pilots in the racing footage come from the newest completed reinforcement-learning checkpoint available when capture preparation began on October 25. It's an experimental snapshot from the current pace-training work, using simulation revision 48, selected to show where training had reached that day. Its promotion status wasn't a condition for appearing in the video.\n\nAll eight racers in each recorded field use that same saved policy. Training can keep moving while the recording holds onto a particular version of the driving behavior.\n\nThat's part of what makes these clips interesting to watch. The ships are working their way through traffic using learned controls, while the replay camera follows the action from rear-quarter, chase and cockpit views.\n\nBehind the footage, this week's RL changes touched difficulty objectives, rewards, checkpoint retention and continuing from earlier training runs. Keeping those snapshots around makes it easier to compare how a change actually affects the driving. The [RL project page](https://projectable.me/projects/nullg/reinforcement-learning/) explains how the policy turns its view of a race into steering, throttle, airbrake, pitch and weapon inputs.\n\n## Shields deserve a close-up\n\nThe shield is one of the highlights of this week's XL effects work. Its animated surface wraps around the hull, with the ship and engine flare still visible through it. It's especially satisfying in a held shot, where there's time to see the surface move.\n\n![Animated blue shield surrounding Qirex in a fresh native XL harness capture.](https://projectable.me/content/images/2026/10/ecc5894c0244-nullg-weekly-fresh-shield-closeup.png)\n\n*Qirex's shield on Sagarmatha, stepped through the native weapon harness.*\n\nThe rest of the arsenal gets plenty of attention too. **Rockets, missile and electrobolt** show their launches, flight and impacts. **Mines, quake, plasma and energy bomb** bring the larger effects into the track view. **Turbo, autopilot and the energy pack** round out the eleven types, alongside the shield, with their activation and status feedback.\n\n![Qirex firing rockets on Sagarmatha in a fresh native weapon harness capture.](https://projectable.me/content/images/2026/10/e4a12b0e8f41-nullg-weekly-fresh-xl-rocket-0.png)\n\n*Rockets leaving Qirex on Sagarmatha, captured from the paused production race harness.*\n\nThese are staged demonstrations using granted inventory, so each weapon gets a clear moment on screen. The mine sequence also places a live mine to show the impact. That keeps the edit moving and makes it possible to spend some time with effects that would otherwise flash past during a race.\n\nThe smaller presentation changes matter here too: targeting trails, status placement and the order effects are drawn all help the action read when several things happen at once.\n\n## A little more room for the camera\n\nGetting this much work into a video also gave the replay tools a workout. Authored camera sequences bring the chase, cockpit and wider angles together, while XL camera handoffs and the results view received stabilization fixes.\n\nUnderneath it all, asset preparation and texture-residency improvements support the richer models. Shared lighting and effects keep the ships looking consistent between menus, inspection views and races. For the engine details, the [architecture page](https://projectable.me/projects/nullg/architecture/) connects those systems to the simulation and asset loading.\n\n## Split-screen gets some breathing room\n\nSplit-screen had a good week too. Each player's HUD now fits the shape of their own viewport, with the top readouts at the top, speed and thrust anchored toward the bottom, and status information kept around the edges. That gives the road ahead more room in both horizontal and vertical layouts.\n\n![Two-player XL split-screen showing independent weapon and shield status in a controlled HUD review.](https://projectable.me/content/images/2026/10/169198aae575-nullg-weekly-fresh-split-hud.png)\n\n*Two-player capture from the native HUD review harness. Weapon, shield and lap states are staged to check how the readouts fit; this is a presentation test.*\n\nThe typography has had attention as well. HD2097 gives the racing labels and large figures their industrial character, while colored status labels and brighter numbers make shield countdowns, health, boost and thrust easier to pick out. Compact text keeps driver names and results readable in the smaller views. Targeting trails and weapon feedback remain tied to the relevant player's camera.\n\nThree-player races make particularly good use of the screen. The fourth quadrant now carries a live race overview: a circuit map with numbered ship markers beside the standings, lap and speed information. Local players keep the same colors across the map, legend and standings, so it's easy to follow who's where.\n\n![Three-player XL split-screen with a live circuit map and standings in the fourth quadrant.](https://projectable.me/content/images/2026/10/cf608e45de29-nullg-weekly-fresh-split-overview.png)\n\n*Three race views and a live overview, captured in the native split-screen harness.*\n\nIt's a nice addition for everyone watching as well as everyone driving. Finished and retired racers get their own result states, and each driver's results stay inside their view. The [multiplayer project page](https://projectable.me/projects/nullg/multiplayer/) covers the wider local and network racing setup.\n\n## Pulse is on the grid\n\nAway from the XL footage, **WipEout Pulse** reached a particularly exciting point: a playable preview through NullG's normal import and menu flow.\n\nThe current slice is **Talon's Junction White, FEISAR and Venom**, with a one-lap Time Trial or a Single Race against seven native opponents. Difficulty selection is available in Race Setup.\n\n![FEISAR with an active Pulse shield on Talon's Junction White in a new native capture.](https://projectable.me/content/images/2026/10/e14ca56845db-nullg-weekly-fresh-pulse-shield.png)\n\n*Pulse in NullG's race scene, with FEISAR's Shield and its HUD countdown visible.*\n\nThere's a lot wrapped up in getting that first race to feel right: grid placement, the countdown, launch behavior, steering, airbrakes and the sideshift gesture. Banking and barrel-roll visuals, a chase camera that follows the track horizon, and feedback for landings, wall contact and recovery bring more of the experience together.\n\nTime Trial now hands out a **Turbo** at the start/lap line. It uses the normal weapon controls, including the option to absorb it for energy, and appears in the common HUD. Ship and track contact also feed into the shared damage handling.\n\nPulse's **Shield and three-shot Rocket bursts** have reached the controlled integration stage too, connecting native weapon state to the model, rendering and HUD systems. The recorded checks exercised shielded and unshielded encounters, and all eight racers completed a lap in the full validation run. These weapon shots use explicit inspection grants to show the new behavior clearly.\n\n![Three Pulse Rocket darts flying ahead of FEISAR on Talon's Junction White.](https://projectable.me/content/images/2026/10/9899d3185aba-nullg-weekly-fresh-pulse-rockets.png)\n\n*Three Rocket darts heading down Talon's Junction White ahead of FEISAR.*\n\n## Pure is moving along with it\n\nPure's progress this week is in the import pipeline: getting its tracks, ships and resources decoded and checked. The shared PSP work covers geometry, materials, textures, collision and animation data, while keeping each game's content distinct.\n\nThe CPU loader checks now cover **16 Pure track layouts and 24 Pulse layouts**. Ship metadata accounts for **eleven Pure ships and eight Pulse ships**, including the resources they depend on. Sound-bank voices and soundtrack resources have also been decoded and checked.\n\nThose numbers describe the content-processing work; Pulse is the current playable preview. Together, they give a good sense of how much of the PSP library is moving through the engine's import tools.\n\nThat work feeds into the existing model cache, staged GPU uploads and track streaming. Pulse's supported disc import validates a private installation before making it active in the asset library, connecting the source data to the same scene and inspection tools used elsewhere in NullG.\n\n## The tools are part of the story\n\nA lot of this week's progress happened with the engine paused and a model, track section or weapon effect under inspection. NullG's MCP harnesses let an agent or script operate the real native tools: load an asset, move the camera, pose the control surfaces, inspect the state and capture the result.\n\nThat makes for a very hands-on development loop. With the XL fleet, the work moves from Blender into the engine's model viewer, through material and motion checks, then into the menu and a race. A beautiful model still needs to look right under the engine's lighting and move cleanly when the airbrakes come out.\n\n![The native Model Viewer inspecting the rebuilt Qirex with bounds and control-surface hierarchy visible.](https://projectable.me/content/images/2026/10/e00793761eb9-nullg-weekly-fresh-mcp-qirex-surfaces.png)\n\n*Qirex in an MCP-driven Model Viewer session. The hierarchy, texture panel and bounds put the rendered ship beside the data being reviewed.*\n\nThe weapon harness is especially handy. It opens a paused production race, grants a selected weapon and lets the simulation advance a few frames at a time. A snapshot can preserve the setup before a launch, so the same moment can be restored and tried again. Ship state, weapon timers and events sit alongside the rendered frames. That's how an effect can get a proper close-up while still going through the game's firing and rendering paths.\n\nThis week also added more targeted Pulse feedback checks. One probe starts a ship above the track with downward velocity, letting the normal hover and collision systems produce the landing. Another moves it beyond the track so the recovery controller has to respond. The reports retain when the feedback happened, which racer produced it and its magnitude. It's a useful way to connect something felt during play to a repeatable case.\n\n![The native Track Validator showing track geometry and inspection controls during a fresh capture session.](https://projectable.me/content/images/2026/10/a43258e42685-nullg-weekly-fresh-mcp-track-inspection.png)\n\n*The Track Validator: track context, geometry selection and texture inspection in one working view.*\n\nThe capture workflow has grown along with the inspection tools. This week's authored replay cameras and music-video tooling turn saved races into planned shots, then into landscape and portrait edits. The racing stays in the replay; the camera and edit can be worked on separately. For Shields Up, that meant retaining the checkpoint, recording each ship's race and staging the weapon demonstrations before assembling the full-song film and short.\n\nThe practical change is how closely those steps now connect. A model review, a controlled weapon test and a finished video can all come out of the same engine tools, with the inputs and reports kept alongside the images. The [MCP tooling page](https://projectable.me/projects/nullg/mcp-tooling/) has a deeper look at the harnesses and how they fit together.\n\n**Warehouse Rage** ties the video together, while playlist playback, track announcements, **Zero G Chokehold** and **On Fire** were also part of this week's music work. The original NullG soundtrack is generated with Suno.\n\nA rebuilt fleet, learned pilots on track, a proper look at the XL arsenal, and Pulse lining up on the grid: it's been a good week to point a camera at NullG. [Watch the showcase](https://www.youtube.com/watch?v=2VYHD8xlsoM), or spend a little longer with the [project pages](https://projectable.me/projects/nullg/) for the work behind the footage.\n\n*WipEout names, designs and imported game content belong to their respective owners. The images show that content and its reconstructed presentation inside NullG.*\n","html":"\u003cp\u003e\u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG\u003c/a\u003e had a busy week, and the best way to catch up is to see it moving. Five rebuilt XL ships, a fresh batch of learned pilots, shields rippling around the hulls, and a playable slice of WipEout Pulse: there\u0026#39;s a lot to get into.\u003c/p\u003e\n\u003cp\u003eThe new \u003cstrong\u003eShields Up\u003c/strong\u003e video pulls the XL work together over the full \u003cstrong\u003eWarehouse Rage\u003c/strong\u003e track. Every ship gets screen time, every weapon gets a turn, and the camera gets close enough to enjoy the details. Turn the sound up for this one.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/2VYHD8xlsoM\" title=\"NullG — Shields Up: weekly development showcase\" loading=\"lazy\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003ca href=\"https://www.youtube.com/watch?v=2VYHD8xlsoM\" rel=\"nofollow\"\u003eWatch Shields Up on YouTube\u003c/a\u003e. It\u0026#39;s 3 minutes and 19 seconds of this week\u0026#39;s work, with a separate 90-second portrait edit for a quicker look.\u003c/p\u003e\n\u003ch2\u003eThe fleet is coming together\u003c/h2\u003e\n\u003cp\u003e\u003cstrong\u003eFEISAR, AG Systems, Auricom, Qirex and Piranha\u003c/strong\u003e all feature in the updated XL fleet. The familiar shapes are there, now with rebuilt geometry, physically based materials, and more attention to the cockpits, engines and moving control surfaces.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/0f9bf03cf410-nullg-weekly-fresh-fleet-a.png\" alt=\"FEISAR and AG Systems in fresh native model-viewer captures.\"\u003e\n\u003cp\u003e\u003cem\u003eFEISAR and AG Systems in the native Model Viewer.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThere\u0026#39;s a lovely amount of character in these ships. Qirex\u0026#39;s broad shape and Piranha\u0026#39;s narrow profile ask very different things of the model work. Even the ailerons need care: their position, thickness and range of movement all affect how convincing the ship looks when it banks into a corner.\u003c/p\u003e\n\u003cp\u003eThis week brought more of those details together, including clearance checks on Qirex and refinements to the thin control surfaces on AG Systems and Piranha. Lighting and materials carry the work from the menus onto the track, through bright open sections, tunnels and city lights.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/a300c3095728-nullg-weekly-fresh-fleet-b.png\" alt=\"Auricom, Qirex and Piranha in fresh native model-viewer captures.\"\u003e\n\u003cp\u003e\u003cem\u003eAuricom, Qirex and Piranha, with control surfaces posed in the engine viewer.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe \u003ca href=\"https://projectable.me/nullg-recreating-wipeout-xl-models-in-blender/\" rel=\"nofollow\"\u003eBlender fleet article\u003c/a\u003e gets into the reconstruction process. The video is the payoff: all five ships, moving through the world they were built for.\u003c/p\u003e\n\u003ch2\u003eFresh pilots, straight from training\u003c/h2\u003e\n\u003cp\u003eThe pilots in the racing footage come from the newest completed reinforcement-learning checkpoint available when capture preparation began on October 25. It\u0026#39;s an experimental snapshot from the current pace-training work, using simulation revision 48, selected to show where training had reached that day. Its promotion status wasn\u0026#39;t a condition for appearing in the video.\u003c/p\u003e\n\u003cp\u003eAll eight racers in each recorded field use that same saved policy. Training can keep moving while the recording holds onto a particular version of the driving behavior.\u003c/p\u003e\n\u003cp\u003eThat\u0026#39;s part of what makes these clips interesting to watch. The ships are working their way through traffic using learned controls, while the replay camera follows the action from rear-quarter, chase and cockpit views.\u003c/p\u003e\n\u003cp\u003eBehind the footage, this week\u0026#39;s RL changes touched difficulty objectives, rewards, checkpoint retention and continuing from earlier training runs. Keeping those snapshots around makes it easier to compare how a change actually affects the driving. The \u003ca href=\"https://projectable.me/projects/nullg/reinforcement-learning/\" rel=\"nofollow\"\u003eRL project page\u003c/a\u003e explains how the policy turns its view of a race into steering, throttle, airbrake, pitch and weapon inputs.\u003c/p\u003e\n\u003ch2\u003eShields deserve a close-up\u003c/h2\u003e\n\u003cp\u003eThe shield is one of the highlights of this week\u0026#39;s XL effects work. Its animated surface wraps around the hull, with the ship and engine flare still visible through it. It\u0026#39;s especially satisfying in a held shot, where there\u0026#39;s time to see the surface move.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/ecc5894c0244-nullg-weekly-fresh-shield-closeup.png\" alt=\"Animated blue shield surrounding Qirex in a fresh native XL harness capture.\"\u003e\n\u003cp\u003e\u003cem\u003eQirex\u0026#39;s shield on Sagarmatha, stepped through the native weapon harness.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe rest of the arsenal gets plenty of attention too. \u003cstrong\u003eRockets, missile and electrobolt\u003c/strong\u003e show their launches, flight and impacts. \u003cstrong\u003eMines, quake, plasma and energy bomb\u003c/strong\u003e bring the larger effects into the track view. \u003cstrong\u003eTurbo, autopilot and the energy pack\u003c/strong\u003e round out the eleven types, alongside the shield, with their activation and status feedback.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/e4a12b0e8f41-nullg-weekly-fresh-xl-rocket-0.png\" alt=\"Qirex firing rockets on Sagarmatha in a fresh native weapon harness capture.\"\u003e\n\u003cp\u003e\u003cem\u003eRockets leaving Qirex on Sagarmatha, captured from the paused production race harness.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThese are staged demonstrations using granted inventory, so each weapon gets a clear moment on screen. The mine sequence also places a live mine to show the impact. That keeps the edit moving and makes it possible to spend some time with effects that would otherwise flash past during a race.\u003c/p\u003e\n\u003cp\u003eThe smaller presentation changes matter here too: targeting trails, status placement and the order effects are drawn all help the action read when several things happen at once.\u003c/p\u003e\n\u003ch2\u003eA little more room for the camera\u003c/h2\u003e\n\u003cp\u003eGetting this much work into a video also gave the replay tools a workout. Authored camera sequences bring the chase, cockpit and wider angles together, while XL camera handoffs and the results view received stabilization fixes.\u003c/p\u003e\n\u003cp\u003eUnderneath it all, asset preparation and texture-residency improvements support the richer models. Shared lighting and effects keep the ships looking consistent between menus, inspection views and races. For the engine details, the \u003ca href=\"https://projectable.me/projects/nullg/architecture/\" rel=\"nofollow\"\u003earchitecture page\u003c/a\u003e connects those systems to the simulation and asset loading.\u003c/p\u003e\n\u003ch2\u003eSplit-screen gets some breathing room\u003c/h2\u003e\n\u003cp\u003eSplit-screen had a good week too. Each player\u0026#39;s HUD now fits the shape of their own viewport, with the top readouts at the top, speed and thrust anchored toward the bottom, and status information kept around the edges. That gives the road ahead more room in both horizontal and vertical layouts.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/169198aae575-nullg-weekly-fresh-split-hud.png\" alt=\"Two-player XL split-screen showing independent weapon and shield status in a controlled HUD review.\"\u003e\n\u003cp\u003e\u003cem\u003eTwo-player capture from the native HUD review harness. Weapon, shield and lap states are staged to check how the readouts fit; this is a presentation test.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe typography has had attention as well. HD2097 gives the racing labels and large figures their industrial character, while colored status labels and brighter numbers make shield countdowns, health, boost and thrust easier to pick out. Compact text keeps driver names and results readable in the smaller views. Targeting trails and weapon feedback remain tied to the relevant player\u0026#39;s camera.\u003c/p\u003e\n\u003cp\u003eThree-player races make particularly good use of the screen. The fourth quadrant now carries a live race overview: a circuit map with numbered ship markers beside the standings, lap and speed information. Local players keep the same colors across the map, legend and standings, so it\u0026#39;s easy to follow who\u0026#39;s where.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/cf608e45de29-nullg-weekly-fresh-split-overview.png\" alt=\"Three-player XL split-screen with a live circuit map and standings in the fourth quadrant.\"\u003e\n\u003cp\u003e\u003cem\u003eThree race views and a live overview, captured in the native split-screen harness.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eIt\u0026#39;s a nice addition for everyone watching as well as everyone driving. Finished and retired racers get their own result states, and each driver\u0026#39;s results stay inside their view. The \u003ca href=\"https://projectable.me/projects/nullg/multiplayer/\" rel=\"nofollow\"\u003emultiplayer project page\u003c/a\u003e covers the wider local and network racing setup.\u003c/p\u003e\n\u003ch2\u003ePulse is on the grid\u003c/h2\u003e\n\u003cp\u003eAway from the XL footage, \u003cstrong\u003eWipEout Pulse\u003c/strong\u003e reached a particularly exciting point: a playable preview through NullG\u0026#39;s normal import and menu flow.\u003c/p\u003e\n\u003cp\u003eThe current slice is \u003cstrong\u003eTalon\u0026#39;s Junction White, FEISAR and Venom\u003c/strong\u003e, with a one-lap Time Trial or a Single Race against seven native opponents. Difficulty selection is available in Race Setup.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/e14ca56845db-nullg-weekly-fresh-pulse-shield.png\" alt=\"FEISAR with an active Pulse shield on Talon\u0026#39;s Junction White in a new native capture.\"\u003e\n\u003cp\u003e\u003cem\u003ePulse in NullG\u0026#39;s race scene, with FEISAR\u0026#39;s Shield and its HUD countdown visible.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThere\u0026#39;s a lot wrapped up in getting that first race to feel right: grid placement, the countdown, launch behavior, steering, airbrakes and the sideshift gesture. Banking and barrel-roll visuals, a chase camera that follows the track horizon, and feedback for landings, wall contact and recovery bring more of the experience together.\u003c/p\u003e\n\u003cp\u003eTime Trial now hands out a \u003cstrong\u003eTurbo\u003c/strong\u003e at the start/lap line. It uses the normal weapon controls, including the option to absorb it for energy, and appears in the common HUD. Ship and track contact also feed into the shared damage handling.\u003c/p\u003e\n\u003cp\u003ePulse\u0026#39;s \u003cstrong\u003eShield and three-shot Rocket bursts\u003c/strong\u003e have reached the controlled integration stage too, connecting native weapon state to the model, rendering and HUD systems. The recorded checks exercised shielded and unshielded encounters, and all eight racers completed a lap in the full validation run. These weapon shots use explicit inspection grants to show the new behavior clearly.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/9899d3185aba-nullg-weekly-fresh-pulse-rockets.png\" alt=\"Three Pulse Rocket darts flying ahead of FEISAR on Talon\u0026#39;s Junction White.\"\u003e\n\u003cp\u003e\u003cem\u003eThree Rocket darts heading down Talon\u0026#39;s Junction White ahead of FEISAR.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003ePure is moving along with it\u003c/h2\u003e\n\u003cp\u003ePure\u0026#39;s progress this week is in the import pipeline: getting its tracks, ships and resources decoded and checked. The shared PSP work covers geometry, materials, textures, collision and animation data, while keeping each game\u0026#39;s content distinct.\u003c/p\u003e\n\u003cp\u003eThe CPU loader checks now cover \u003cstrong\u003e16 Pure track layouts and 24 Pulse layouts\u003c/strong\u003e. Ship metadata accounts for \u003cstrong\u003eeleven Pure ships and eight Pulse ships\u003c/strong\u003e, including the resources they depend on. Sound-bank voices and soundtrack resources have also been decoded and checked.\u003c/p\u003e\n\u003cp\u003eThose numbers describe the content-processing work; Pulse is the current playable preview. Together, they give a good sense of how much of the PSP library is moving through the engine\u0026#39;s import tools.\u003c/p\u003e\n\u003cp\u003eThat work feeds into the existing model cache, staged GPU uploads and track streaming. Pulse\u0026#39;s supported disc import validates a private installation before making it active in the asset library, connecting the source data to the same scene and inspection tools used elsewhere in NullG.\u003c/p\u003e\n\u003ch2\u003eThe tools are part of the story\u003c/h2\u003e\n\u003cp\u003eA lot of this week\u0026#39;s progress happened with the engine paused and a model, track section or weapon effect under inspection. NullG\u0026#39;s MCP harnesses let an agent or script operate the real native tools: load an asset, move the camera, pose the control surfaces, inspect the state and capture the result.\u003c/p\u003e\n\u003cp\u003eThat makes for a very hands-on development loop. With the XL fleet, the work moves from Blender into the engine\u0026#39;s model viewer, through material and motion checks, then into the menu and a race. A beautiful model still needs to look right under the engine\u0026#39;s lighting and move cleanly when the airbrakes come out.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/e00793761eb9-nullg-weekly-fresh-mcp-qirex-surfaces.png\" alt=\"The native Model Viewer inspecting the rebuilt Qirex with bounds and control-surface hierarchy visible.\"\u003e\n\u003cp\u003e\u003cem\u003eQirex in an MCP-driven Model Viewer session. The hierarchy, texture panel and bounds put the rendered ship beside the data being reviewed.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe weapon harness is especially handy. It opens a paused production race, grants a selected weapon and lets the simulation advance a few frames at a time. A snapshot can preserve the setup before a launch, so the same moment can be restored and tried again. Ship state, weapon timers and events sit alongside the rendered frames. That\u0026#39;s how an effect can get a proper close-up while still going through the game\u0026#39;s firing and rendering paths.\u003c/p\u003e\n\u003cp\u003eThis week also added more targeted Pulse feedback checks. One probe starts a ship above the track with downward velocity, letting the normal hover and collision systems produce the landing. Another moves it beyond the track so the recovery controller has to respond. The reports retain when the feedback happened, which racer produced it and its magnitude. It\u0026#39;s a useful way to connect something felt during play to a repeatable case.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/a43258e42685-nullg-weekly-fresh-mcp-track-inspection.png\" alt=\"The native Track Validator showing track geometry and inspection controls during a fresh capture session.\"\u003e\n\u003cp\u003e\u003cem\u003eThe Track Validator: track context, geometry selection and texture inspection in one working view.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe capture workflow has grown along with the inspection tools. This week\u0026#39;s authored replay cameras and music-video tooling turn saved races into planned shots, then into landscape and portrait edits. The racing stays in the replay; the camera and edit can be worked on separately. For Shields Up, that meant retaining the checkpoint, recording each ship\u0026#39;s race and staging the weapon demonstrations before assembling the full-song film and short.\u003c/p\u003e\n\u003cp\u003eThe practical change is how closely those steps now connect. A model review, a controlled weapon test and a finished video can all come out of the same engine tools, with the inputs and reports kept alongside the images. The \u003ca href=\"https://projectable.me/projects/nullg/mcp-tooling/\" rel=\"nofollow\"\u003eMCP tooling page\u003c/a\u003e has a deeper look at the harnesses and how they fit together.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWarehouse Rage\u003c/strong\u003e ties the video together, while playlist playback, track announcements, \u003cstrong\u003eZero G Chokehold\u003c/strong\u003e and \u003cstrong\u003eOn Fire\u003c/strong\u003e were also part of this week\u0026#39;s music work. The original NullG soundtrack is generated with Suno.\u003c/p\u003e\n\u003cp\u003eA rebuilt fleet, learned pilots on track, a proper look at the XL arsenal, and Pulse lining up on the grid: it\u0026#39;s been a good week to point a camera at NullG. \u003ca href=\"https://www.youtube.com/watch?v=2VYHD8xlsoM\" rel=\"nofollow\"\u003eWatch the showcase\u003c/a\u003e, or spend a little longer with the \u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eproject pages\u003c/a\u003e for the work behind the footage.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eWipEout names, designs and imported game content belong to their respective owners. The images show that content and its reconstructed presentation inside NullG.\u003c/em\u003e\u003c/p\u003e\n","excerpt":"A busy week of rebuilt XL ships, learned pilots, shields and split-screen polish, with Pulse on the grid and a closer look at the MCP tools behind it.","metaTitle":"NullG weekly update: ships, shields, Pulse and MCP tools","metaDescription":"New XL ships, experimental RL racing, weapon effects, split-screen HUD improvements, Pulse and Pure progress, and the MCP harnesses behind this week’s work.","status":"published","featured":false,"publishedAt":"2026-10-25T23:34:35Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"b6d17adfd043ab59aaf5a462247eea23","projectId":"ae4bdb66da6dc1729a263a475e1cdbe3","tags":[{"id":"10a974c0a281d4fcf3d44f5ef8dae079","name":"NullG","slug":"nullg"},{"id":"1c9e8a1339a8f8b797767e91a9673016","name":"Game","slug":"game"},{"id":"60103f47eb65b46c6138b9d6fc7b17d0","name":"Machine Learning","slug":"machine-learning"}],"createdAt":"2026-10-25T23:12:18Z","updatedAt":"2026-10-25T23:34:35Z"},{"id":"9f1711a950f76598b42e3b1fd15b45f0","siteId":"projectable","slug":"nullg-recreating-wipeout-xl-models-in-blender","title":"Recreating the WipEout XL fleet in Blender","markdown":"[NullG](https://projectable.me/projects/nullg/)’s WipEout XL fleet now has a complete set of Blender-authored recreations: Feisar, Qirex, Auricom, AG Systems, and Piranha. Their familiar silhouettes and team colors carry through into inset cockpit glass, working control surfaces, worn vinyl graphics, and detailed engine housings.\n\nThe same process has also reached the smaller objects that give a race its character: missiles, rockets, mines, the recovery drone, starting lights, and track cameras. Together, they show how careful reference work, practical modeling constraints, and repeatable visual reviews can bring a consistent finish to a whole collection.\n\n## Start with the shape everyone recognizes\n\nFeisar was the pilot for the workflow. Original XL geometry and texture placement provided the measurements: overall proportions, wing profiles, cockpit position, engine openings, and the blue-and-yellow livery. Front, rear, side, top, and underside comparisons gave each modeling decision a clear reference.\n\nThe finished surfaces were rebuilt in Blender, with the source available as a measurement reference. Controlled bevels add definition to edges. Recessed engine throats, seated fairings, and panel thickness give the hull a physical structure. Close-up views establish how those pieces meet; the overall views keep the ship recognizable at racing distance.\n\nEach review focused on a manageable stage. Shape came first, followed by cockpit and engine details, then paint and wear, articulation, and native rendering. Feedback became concrete adjustments: a windshield corner radius, a shared window boundary, a cleaner emblem outline, or a thinner moving surface.\n\n## Blender, connected to the review loop\n\nBlender’s community MCP integration gives the assistant a direct connection to the authoring scene. MCP—the Model Context Protocol—exposes tools for inspecting scene contents, running scoped Blender operations, and reviewing the viewport. In NullG’s setup, those operations happen in a dedicated Blender session with a saved revision to work from.\n\nInteractive inspection sits alongside scripted Blender recipes. The scripts make geometry construction, material setup, export, and evidence capture repeatable. An editable `.blend` retains the authoring work, while a self-contained GLB carries meshes, materials, textures, and named parts into the engine.\n\nThe exported GLB gets its own review. It’s imported into a clean Blender scene so the cameras see the actual delivery file, including the hierarchy, glass, and embedded textures. Numbered revisions and recorded asset hashes keep the screenshots connected to the exact model under discussion.\n\nThat combination makes iteration particularly useful: a small, specific visual request can become an editable rule, then be checked again from the same angles.\n\n## Cockpits built from shared constraints\n\nThe cockpit work established one of the most reusable patterns in the project. Each window, seal, and recess derives from the same opening in the canopy shell. Frame margins and inset depths are explicit. Neighboring panes share alignment landmarks, so a change to the canopy has a clear relationship to the surrounding glass.\n\nThis gives each team its own window shapes while keeping the construction consistent. Feisar’s separate panes follow its angular canopy. Auricom’s forward glazing and roof window sit within the red cockpit section. Piranha’s side windows use a deliberately refined profile, with an even orange surround and a flat lower edge.\n\nThe same idea applies elsewhere: vents follow the panel they sit inside, decals use clean outlines, and lens rims meet their housings at a defined depth. Small relationships like these make close-up views feel assembled and intentional.\n\n## Five ships, five distinct finishes\n\nThe fleet shares a material language: lightly glossy painted panels, exposed alloy in contact scratches, darker machinery, and tinted glass. Each ship keeps its own livery and construction details.\n\n**Feisar** established the finish. Its yellow nose carries heavier contact wear, the blue wings retain clear color blocks, and the emblem reads as a worn vinyl sticker over paint. Underneath, a broad abrasion area sits toward the nose, following the idea of a hull that has brushed the track.\n\n![Feisar overall, cockpit, engine and underside Blender review](https://projectable.me/content/images/2026/10/459d0ad8b8c6-nullg-blender-feisar.png)\n\n*Feisar’s cockpit margins, recessed engine, and forward underside abrasion in the Blender review captures.*\n\n**Qirex** carries that approach across a very different hull. Its long purple forms frame a central cockpit, with mechanical texture continuing across the rear and into the engine. The emblem’s repeated slots use consistent dimensions, while surface wear sits mainly on exposed outer areas.\n\n![Qirex overall, cockpit, engine and underside Blender review](https://projectable.me/content/images/2026/10/050f3aa28f05-nullg-blender-qirex.png)\n\n*Qirex’s twin-sided hull, aligned glazing, and rear mechanical surfaces.*\n\n**Auricom** brings a sharper canopy arrangement and a strong red, blue, and dark-panel palette. The front vents follow the windshield contour with a consistent margin. Clean star-shaped artwork and deliberate engine texture coverage keep the livery readable from both ends of the ship.\n\n![Auricom overall, cockpit, engine and underside Blender review](https://projectable.me/content/images/2026/10/158781f9576f-nullg-blender-auricom.png)\n\n*Auricom’s forward glazing, roof window, wing graphics, and engine treatment.*\n\n**AG Systems** uses crisp red-and-white divisions and carefully placed lettering. Its cockpit-adjacent vents follow their supporting panels, and the rear control surfaces have a thin profile that suits the wing geometry. Wear adds contact history along exposed areas while leaving the main graphics easy to read.\n\n![Ag Systems overall, cockpit, engine and underside Blender review](https://projectable.me/content/images/2026/10/7b3dc657056f-nullg-blender-ag_systems.png)\n\n*AG Systems: panel-aligned vents, red-and-white paint, and a detailed rear assembly.*\n\n**Piranha** pairs warm red wings with a continuous orange canopy section. The crown-like rear motif and elongated diamond cutouts were reconstructed with clean lines. The paint wraps around the glazing and across adjoining faces, with the engine and rear wing finish following the physical components.\n\n![Piranha overall, cockpit, engine and underside Blender review](https://projectable.me/content/images/2026/10/da4669288dff-nullg-blender-piranha.png)\n\n*Piranha’s orange canopy surround, reconstructed top motif, and rear wing and engine detail.*\n\n## Wear that follows how a machine gets used\n\nThe texture work became a process of assigning causes to wear. Leading edges receive directional scuffs. Selected seam sections have softly blended abrasion. Engine housings and service areas carry denser mechanical aging. Broad areas of intact paint give those details room to read.\n\nLeft and right sides get independent wear placement. Separate masks control paint abrasion, sticker damage, engine grime, and exposed metal, making each layer easy to adjust. Vinyl chips reveal the paint beneath; deeper hull scratches reveal alloy. The recovery drone adds subtle oxidation appropriate to a well-used piece of track equipment.\n\nBase color establishes the paint and markings. Roughness controls the width and softness of reflections, metallic values distinguish exposed alloy, and normal maps give shallow seams, rivets, and fasteners their surface relief. Together, those maps preserve a gentle paint sheen and brighter metallic scratches, with a rougher finish under the hull and around machinery.\n\nShip texture masters stay at 4K, with runtime texture tiers prepared separately. Lettering uses the appropriate game font where available, including the XL/2097 font for the drone. Logos and cutouts receive their own deliberate tracing and spacing checks.\n\n## Giving the smaller models the same attention\n\nWeapons use a compact geometry budget, with detail concentrated where it changes the outline or material response. The accepted missile has 352 triangles, the rocket 294, and the mine 48. XL supplies their shape and colors; Pulse provides a secondary reference for mechanical detailing.\n\n![Low-poly missile, rocket and mine Blender renders](https://projectable.me/content/images/2026/10/6bf9b9d401b7-nullg-blender-weapons.png)\n\n*The missile, rocket, and mine in Blender review renders. Paint, exposed metal, and local wear provide detail within small meshes.*\n\n![Animated native mine model review](https://projectable.me/content/images/2026/10/c05ec115e332-nullg-blender-mine-animation.png)\n\n*Animated PNG from the native model harness, reviewing the mine in motion.*\n\nThe recovery drone brings more surfaces into play: worn grey panels, rivets, continuous hazard stripes, a centered rear warning label, and modeled front and rear lamps. Its accepted mesh has 638 triangles. Clearer glass response and striped lamp surfaces give the lights depth, while the housing carries the age and use expected of trackside equipment.\n\n![Recovery drone panel, lamp, lettering and underside details](https://projectable.me/content/images/2026/10/e52fe78331da-nullg-blender-drone.png)\n\n*Recovery drone details: hazard striping, mechanical panels, lamp geometry, and the rear warning text.*\n\n![Animated rear lights on the recovery drone](https://projectable.me/content/images/2026/10/5bcccb6c1d7f-nullg-blender-drone-animation.png)\n\n*The drone’s rear lights in the native harness. This animated PNG preserves the reviewed lamp sequence.*\n\nThe starting-light assembly follows the same approach at 308 triangles. Edge wear and grime give the frame an older industrial finish, and recessed glass sections carry a fine line pattern. The harness can show the sequence directly, making the relationship between glass, light, and housing easy to review.\n\n![Animated XL starting-light sequence](https://projectable.me/content/images/2026/10/da33d7093b39-nullg-blender-start-animation.png)\n\n*The starting lights moving through their sequence in the native model harness.*\n\nTrack cameras round out the collection with yellow-and-black stripes, worn grey panels, and clear recessed optics. Their flared lens surrounds and stronger inner bevels give the front a precise mechanical shape. The final seating pass brings the lenses into contact with their recesses. Gameplay validation then checks the replacement across all six XL tracks that contain these cameras, preserving the source placement and animation.\n\n![Track camera lens, top panel and rear housing details](https://projectable.me/content/images/2026/10/947bb5416a02-nullg-blender-camera.png)\n\n*The completed camera’s optics, stripe alignment, and worn top and rear panels.*\n\n## Reviewing movement in the engine\n\nNullG’s [model viewers and MCP inspection harnesses](https://projectable.me/projects/nullg/mcp-tooling/) turn the exported assets into repeatable review subjects. Camera presets, material views, and controlled poses make it possible to compare the same component under the same conditions. Steering and individual airbrake inputs exercise the ship parts, with close-up views around hinges and pockets to check clearance.\n\nThe harness also checks exhaust plumes and trails against the model’s attachment locations. Those details matter together: the engine opening, the plume origin, and a moving control surface all need to occupy the right space.\n\n![All five XL ships with blue engine flares, straight and curved exhaust trails, and control-surface poses](https://projectable.me/content/images/2026/10/f22de3d95342-fleet-exhaust-current.png)\n\n*Fresh native harness captures of all five ships with visible engine flares and exhaust trails. Straight exhaust histories are on the left; curved histories with steering and left-airbrake poses are on the right.*\n\nAnimated effects add another useful test. The shield review places swirling plasma around Feisar, showing its scale, transparency, motion, and relationship to the ship in one view. The Blender-authored envelope and the runtime shader each have a clear role in the finished effect.\n\n![Animated blue plasma shield surrounding Feisar](https://projectable.me/content/images/2026/10/3f2525adca53-nullg-blender-shield-animation.png)\n\n*Animated PNG from the shield context review: flowing blue plasma around Feisar, with the ship visible through the effect.*\n\n## From model review to the starting grid\n\nThe final pass brings the assets into menus and races. The [NullG architecture overview](https://projectable.me/projects/nullg/architecture/) explains the engine systems that support the rendering and simulation behind those scenes. Bright track sections, tunnels, and shadows provide different conditions for checking paint sheen, exposed metal, glass, and texture relief. Scale and attachment checks connect the authored model to the existing racing behavior.\n\n![Five modern XL ships in the pre-game menus](https://projectable.me/content/images/2026/10/f81af4e3c125-fleet-menu-current.png)\n\n*All five modern ships in the current pre-game presentation, freshly captured with the production menu lighting.*\n\n![Feisar gameplay captures in varied track lighting](https://projectable.me/content/images/2026/10/491b2039f780-nullg-blender-lighting.png)\n\n*Feisar in bright, enclosed, and shadowed track conditions during gameplay validation.*\n\nThe result is a coherent set of reference-derived WipEout XL recreations with editable Blender sources, compact runtime exports, and a documented review trail. Each ship retains its team identity, and the smaller objects carry the same care through to the track. The process now has a shared vocabulary—window margins, seated parts, readable vinyl, asymmetric contact wear, and repeatable poses—that makes the next modeling conversation much more precise.\n\n\nExplore the [NullG project homepage](https://projectable.me/projects/nullg/) for the broader engine story, project documentation, and further development updates.\n","html":"\u003cp\u003e\u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG\u003c/a\u003e’s WipEout XL fleet now has a complete set of Blender-authored recreations: Feisar, Qirex, Auricom, AG Systems, and Piranha. Their familiar silhouettes and team colors carry through into inset cockpit glass, working control surfaces, worn vinyl graphics, and detailed engine housings.\u003c/p\u003e\n\u003cp\u003eThe same process has also reached the smaller objects that give a race its character: missiles, rockets, mines, the recovery drone, starting lights, and track cameras. Together, they show how careful reference work, practical modeling constraints, and repeatable visual reviews can bring a consistent finish to a whole collection.\u003c/p\u003e\n\u003ch2\u003eStart with the shape everyone recognizes\u003c/h2\u003e\n\u003cp\u003eFeisar was the pilot for the workflow. Original XL geometry and texture placement provided the measurements: overall proportions, wing profiles, cockpit position, engine openings, and the blue-and-yellow livery. Front, rear, side, top, and underside comparisons gave each modeling decision a clear reference.\u003c/p\u003e\n\u003cp\u003eThe finished surfaces were rebuilt in Blender, with the source available as a measurement reference. Controlled bevels add definition to edges. Recessed engine throats, seated fairings, and panel thickness give the hull a physical structure. Close-up views establish how those pieces meet; the overall views keep the ship recognizable at racing distance.\u003c/p\u003e\n\u003cp\u003eEach review focused on a manageable stage. Shape came first, followed by cockpit and engine details, then paint and wear, articulation, and native rendering. Feedback became concrete adjustments: a windshield corner radius, a shared window boundary, a cleaner emblem outline, or a thinner moving surface.\u003c/p\u003e\n\u003ch2\u003eBlender, connected to the review loop\u003c/h2\u003e\n\u003cp\u003eBlender’s community MCP integration gives the assistant a direct connection to the authoring scene. MCP—the Model Context Protocol—exposes tools for inspecting scene contents, running scoped Blender operations, and reviewing the viewport. In NullG’s setup, those operations happen in a dedicated Blender session with a saved revision to work from.\u003c/p\u003e\n\u003cp\u003eInteractive inspection sits alongside scripted Blender recipes. The scripts make geometry construction, material setup, export, and evidence capture repeatable. An editable \u003ccode\u003e.blend\u003c/code\u003e retains the authoring work, while a self-contained GLB carries meshes, materials, textures, and named parts into the engine.\u003c/p\u003e\n\u003cp\u003eThe exported GLB gets its own review. It’s imported into a clean Blender scene so the cameras see the actual delivery file, including the hierarchy, glass, and embedded textures. Numbered revisions and recorded asset hashes keep the screenshots connected to the exact model under discussion.\u003c/p\u003e\n\u003cp\u003eThat combination makes iteration particularly useful: a small, specific visual request can become an editable rule, then be checked again from the same angles.\u003c/p\u003e\n\u003ch2\u003eCockpits built from shared constraints\u003c/h2\u003e\n\u003cp\u003eThe cockpit work established one of the most reusable patterns in the project. Each window, seal, and recess derives from the same opening in the canopy shell. Frame margins and inset depths are explicit. Neighboring panes share alignment landmarks, so a change to the canopy has a clear relationship to the surrounding glass.\u003c/p\u003e\n\u003cp\u003eThis gives each team its own window shapes while keeping the construction consistent. Feisar’s separate panes follow its angular canopy. Auricom’s forward glazing and roof window sit within the red cockpit section. Piranha’s side windows use a deliberately refined profile, with an even orange surround and a flat lower edge.\u003c/p\u003e\n\u003cp\u003eThe same idea applies elsewhere: vents follow the panel they sit inside, decals use clean outlines, and lens rims meet their housings at a defined depth. Small relationships like these make close-up views feel assembled and intentional.\u003c/p\u003e\n\u003ch2\u003eFive ships, five distinct finishes\u003c/h2\u003e\n\u003cp\u003eThe fleet shares a material language: lightly glossy painted panels, exposed alloy in contact scratches, darker machinery, and tinted glass. Each ship keeps its own livery and construction details.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFeisar\u003c/strong\u003e established the finish. Its yellow nose carries heavier contact wear, the blue wings retain clear color blocks, and the emblem reads as a worn vinyl sticker over paint. Underneath, a broad abrasion area sits toward the nose, following the idea of a hull that has brushed the track.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/459d0ad8b8c6-nullg-blender-feisar.png\" alt=\"Feisar overall, cockpit, engine and underside Blender review\"\u003e\n\u003cp\u003e\u003cem\u003eFeisar’s cockpit margins, recessed engine, and forward underside abrasion in the Blender review captures.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQirex\u003c/strong\u003e carries that approach across a very different hull. Its long purple forms frame a central cockpit, with mechanical texture continuing across the rear and into the engine. The emblem’s repeated slots use consistent dimensions, while surface wear sits mainly on exposed outer areas.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/050f3aa28f05-nullg-blender-qirex.png\" alt=\"Qirex overall, cockpit, engine and underside Blender review\"\u003e\n\u003cp\u003e\u003cem\u003eQirex’s twin-sided hull, aligned glazing, and rear mechanical surfaces.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuricom\u003c/strong\u003e brings a sharper canopy arrangement and a strong red, blue, and dark-panel palette. The front vents follow the windshield contour with a consistent margin. Clean star-shaped artwork and deliberate engine texture coverage keep the livery readable from both ends of the ship.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/158781f9576f-nullg-blender-auricom.png\" alt=\"Auricom overall, cockpit, engine and underside Blender review\"\u003e\n\u003cp\u003e\u003cem\u003eAuricom’s forward glazing, roof window, wing graphics, and engine treatment.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAG Systems\u003c/strong\u003e uses crisp red-and-white divisions and carefully placed lettering. Its cockpit-adjacent vents follow their supporting panels, and the rear control surfaces have a thin profile that suits the wing geometry. Wear adds contact history along exposed areas while leaving the main graphics easy to read.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/7b3dc657056f-nullg-blender-ag_systems.png\" alt=\"Ag Systems overall, cockpit, engine and underside Blender review\"\u003e\n\u003cp\u003e\u003cem\u003eAG Systems: panel-aligned vents, red-and-white paint, and a detailed rear assembly.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePiranha\u003c/strong\u003e pairs warm red wings with a continuous orange canopy section. The crown-like rear motif and elongated diamond cutouts were reconstructed with clean lines. The paint wraps around the glazing and across adjoining faces, with the engine and rear wing finish following the physical components.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/da4669288dff-nullg-blender-piranha.png\" alt=\"Piranha overall, cockpit, engine and underside Blender review\"\u003e\n\u003cp\u003e\u003cem\u003ePiranha’s orange canopy surround, reconstructed top motif, and rear wing and engine detail.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eWear that follows how a machine gets used\u003c/h2\u003e\n\u003cp\u003eThe texture work became a process of assigning causes to wear. Leading edges receive directional scuffs. Selected seam sections have softly blended abrasion. Engine housings and service areas carry denser mechanical aging. Broad areas of intact paint give those details room to read.\u003c/p\u003e\n\u003cp\u003eLeft and right sides get independent wear placement. Separate masks control paint abrasion, sticker damage, engine grime, and exposed metal, making each layer easy to adjust. Vinyl chips reveal the paint beneath; deeper hull scratches reveal alloy. The recovery drone adds subtle oxidation appropriate to a well-used piece of track equipment.\u003c/p\u003e\n\u003cp\u003eBase color establishes the paint and markings. Roughness controls the width and softness of reflections, metallic values distinguish exposed alloy, and normal maps give shallow seams, rivets, and fasteners their surface relief. Together, those maps preserve a gentle paint sheen and brighter metallic scratches, with a rougher finish under the hull and around machinery.\u003c/p\u003e\n\u003cp\u003eShip texture masters stay at 4K, with runtime texture tiers prepared separately. Lettering uses the appropriate game font where available, including the XL/2097 font for the drone. Logos and cutouts receive their own deliberate tracing and spacing checks.\u003c/p\u003e\n\u003ch2\u003eGiving the smaller models the same attention\u003c/h2\u003e\n\u003cp\u003eWeapons use a compact geometry budget, with detail concentrated where it changes the outline or material response. The accepted missile has 352 triangles, the rocket 294, and the mine 48. XL supplies their shape and colors; Pulse provides a secondary reference for mechanical detailing.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/6bf9b9d401b7-nullg-blender-weapons.png\" alt=\"Low-poly missile, rocket and mine Blender renders\"\u003e\n\u003cp\u003e\u003cem\u003eThe missile, rocket, and mine in Blender review renders. Paint, exposed metal, and local wear provide detail within small meshes.\u003c/em\u003e\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/c05ec115e332-nullg-blender-mine-animation.png\" alt=\"Animated native mine model review\"\u003e\n\u003cp\u003e\u003cem\u003eAnimated PNG from the native model harness, reviewing the mine in motion.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe recovery drone brings more surfaces into play: worn grey panels, rivets, continuous hazard stripes, a centered rear warning label, and modeled front and rear lamps. Its accepted mesh has 638 triangles. Clearer glass response and striped lamp surfaces give the lights depth, while the housing carries the age and use expected of trackside equipment.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/e52fe78331da-nullg-blender-drone.png\" alt=\"Recovery drone panel, lamp, lettering and underside details\"\u003e\n\u003cp\u003e\u003cem\u003eRecovery drone details: hazard striping, mechanical panels, lamp geometry, and the rear warning text.\u003c/em\u003e\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/5bcccb6c1d7f-nullg-blender-drone-animation.png\" alt=\"Animated rear lights on the recovery drone\"\u003e\n\u003cp\u003e\u003cem\u003eThe drone’s rear lights in the native harness. This animated PNG preserves the reviewed lamp sequence.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe starting-light assembly follows the same approach at 308 triangles. Edge wear and grime give the frame an older industrial finish, and recessed glass sections carry a fine line pattern. The harness can show the sequence directly, making the relationship between glass, light, and housing easy to review.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/da33d7093b39-nullg-blender-start-animation.png\" alt=\"Animated XL starting-light sequence\"\u003e\n\u003cp\u003e\u003cem\u003eThe starting lights moving through their sequence in the native model harness.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTrack cameras round out the collection with yellow-and-black stripes, worn grey panels, and clear recessed optics. Their flared lens surrounds and stronger inner bevels give the front a precise mechanical shape. The final seating pass brings the lenses into contact with their recesses. Gameplay validation then checks the replacement across all six XL tracks that contain these cameras, preserving the source placement and animation.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/947bb5416a02-nullg-blender-camera.png\" alt=\"Track camera lens, top panel and rear housing details\"\u003e\n\u003cp\u003e\u003cem\u003eThe completed camera’s optics, stripe alignment, and worn top and rear panels.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eReviewing movement in the engine\u003c/h2\u003e\n\u003cp\u003eNullG’s \u003ca href=\"https://projectable.me/projects/nullg/mcp-tooling/\" rel=\"nofollow\"\u003emodel viewers and MCP inspection harnesses\u003c/a\u003e turn the exported assets into repeatable review subjects. Camera presets, material views, and controlled poses make it possible to compare the same component under the same conditions. Steering and individual airbrake inputs exercise the ship parts, with close-up views around hinges and pockets to check clearance.\u003c/p\u003e\n\u003cp\u003eThe harness also checks exhaust plumes and trails against the model’s attachment locations. Those details matter together: the engine opening, the plume origin, and a moving control surface all need to occupy the right space.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/f22de3d95342-fleet-exhaust-current.png\" alt=\"All five XL ships with blue engine flares, straight and curved exhaust trails, and control-surface poses\"\u003e\n\u003cp\u003e\u003cem\u003eFresh native harness captures of all five ships with visible engine flares and exhaust trails. Straight exhaust histories are on the left; curved histories with steering and left-airbrake poses are on the right.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eAnimated effects add another useful test. The shield review places swirling plasma around Feisar, showing its scale, transparency, motion, and relationship to the ship in one view. The Blender-authored envelope and the runtime shader each have a clear role in the finished effect.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/3f2525adca53-nullg-blender-shield-animation.png\" alt=\"Animated blue plasma shield surrounding Feisar\"\u003e\n\u003cp\u003e\u003cem\u003eAnimated PNG from the shield context review: flowing blue plasma around Feisar, with the ship visible through the effect.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eFrom model review to the starting grid\u003c/h2\u003e\n\u003cp\u003eThe final pass brings the assets into menus and races. The \u003ca href=\"https://projectable.me/projects/nullg/architecture/\" rel=\"nofollow\"\u003eNullG architecture overview\u003c/a\u003e explains the engine systems that support the rendering and simulation behind those scenes. Bright track sections, tunnels, and shadows provide different conditions for checking paint sheen, exposed metal, glass, and texture relief. Scale and attachment checks connect the authored model to the existing racing behavior.\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/f81af4e3c125-fleet-menu-current.png\" alt=\"Five modern XL ships in the pre-game menus\"\u003e\n\u003cp\u003e\u003cem\u003eAll five modern ships in the current pre-game presentation, freshly captured with the production menu lighting.\u003c/em\u003e\u003c/p\u003e\n\u003cimg src=\"https://projectable.me/content/images/2026/10/491b2039f780-nullg-blender-lighting.png\" alt=\"Feisar gameplay captures in varied track lighting\"\u003e\n\u003cp\u003e\u003cem\u003eFeisar in bright, enclosed, and shadowed track conditions during gameplay validation.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe result is a coherent set of reference-derived WipEout XL recreations with editable Blender sources, compact runtime exports, and a documented review trail. Each ship retains its team identity, and the smaller objects carry the same care through to the track. The process now has a shared vocabulary—window margins, seated parts, readable vinyl, asymmetric contact wear, and repeatable poses—that makes the next modeling conversation much more precise.\u003c/p\u003e\n\u003cp\u003eExplore the \u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG project homepage\u003c/a\u003e for the broader engine story, project documentation, and further development updates.\u003c/p\u003e\n","excerpt":"A closer look at NullG’s recreated XL ships and trackside models, from constrained cockpit glass and worn paint to Blender MCP, animated previews, and in-engine validation.","metaTitle":"Recreating WipEout XL models in Blender | NullG","metaDescription":"Explore NullG’s Blender workflow for five WipEout XL ships, weapons, drones and track cameras, with detailed galleries and animated MCP harness captures.","status":"published","featured":false,"publishedAt":"2026-10-25T14:05:33Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"ec095655019808e4dac9f18d8ff4eb75","projectId":"ae4bdb66da6dc1729a263a475e1cdbe3","tags":[{"id":"1c9e8a1339a8f8b797767e91a9673016","name":"Game","slug":"game"},{"id":"10a974c0a281d4fcf3d44f5ef8dae079","name":"NullG","slug":"nullg"}],"createdAt":"2026-10-25T13:58:21Z","updatedAt":"2026-10-25T14:11:23Z"},{"id":"71c64fb16098ef3fc7196432defc52c7","siteId":"projectable","slug":"nullg-original-soundtrack","title":"A new soundtrack for NullG","markdown":"There’s some new music in [NullG](https://projectable.me/projects/nullg/). Seven instrumental tracks, made with [Suno](https://suno.com/), bring a little 90s drum and bass, jungle and EDM to the racing.\n\nThe inspiration comes from the original WipEout soundtracks, especially CoLD SToRAGE’s work. The aim was to capture some of that feeling in a new collection of songs for NullG. One track, “Cold Storage”, is a small homage to that influence.\n\nThere’s a practical reason for making new music, too. The goal is to build up original and re-imagined content that can be shared with NullG, rather than bundling copyrighted music and assets from the original games. This soundtrack is a start in that direction.\n\nFor now, the songs have their own entries in the NullG Cinematic Music Video series, paired with footage of classic WipEout circuits running in the engine. Each video plays the full track, with just the music to listen to.\n\nHave a listen below, or open the [YouTube playlist](https://www.youtube.com/watch?v=SF1iQXejQEk\u0026list=PLaR-_mRLlb6E) and let it play.\n\n### Zero G Chokehold\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/SF1iQXejQEk\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Talon’s Reach.*\n\n### On Fire\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/pyEm89gGd0E\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Vostok Island.*\n\n### Cold Storage\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/yS-XkE8beWI\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Sagarmatha.*\n\n### Flame Run\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/NWQp7iDyqoY\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Talon’s Reach.*\n\n### Hyperdrive Inferno\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/8dJ4h6yr2No\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Gare d’Europa.*\n\n### Turbo Ignition\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/0axF9qJyirI\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Phenitia Park.*\n\n### Warehouse Rage\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/VmO_zsf-ccY\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Cinematic race footage: Odessa Keys.*\n","html":"\u003cp\u003eThere’s some new music in \u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG\u003c/a\u003e. Seven instrumental tracks, made with \u003ca href=\"https://suno.com/\" rel=\"nofollow\"\u003eSuno\u003c/a\u003e, bring a little 90s drum and bass, jungle and EDM to the racing.\u003c/p\u003e\n\u003cp\u003eThe inspiration comes from the original WipEout soundtracks, especially CoLD SToRAGE’s work. The aim was to capture some of that feeling in a new collection of songs for NullG. One track, “Cold Storage”, is a small homage to that influence.\u003c/p\u003e\n\u003cp\u003eThere’s a practical reason for making new music, too. The goal is to build up original and re-imagined content that can be shared with NullG, rather than bundling copyrighted music and assets from the original games. This soundtrack is a start in that direction.\u003c/p\u003e\n\u003cp\u003eFor now, the songs have their own entries in the NullG Cinematic Music Video series, paired with footage of classic WipEout circuits running in the engine. Each video plays the full track, with just the music to listen to.\u003c/p\u003e\n\u003cp\u003eHave a listen below, or open the \u003ca href=\"https://www.youtube.com/watch?v=SF1iQXejQEk\u0026amp;list=PLaR-_mRLlb6E\" rel=\"nofollow\"\u003eYouTube playlist\u003c/a\u003e and let it play.\u003c/p\u003e\n\u003ch3\u003eZero G Chokehold\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/SF1iQXejQEk\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Talon’s Reach.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eOn Fire\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/pyEm89gGd0E\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Vostok Island.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eCold Storage\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/yS-XkE8beWI\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Sagarmatha.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eFlame Run\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/NWQp7iDyqoY\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Talon’s Reach.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eHyperdrive Inferno\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/8dJ4h6yr2No\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Gare d’Europa.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eTurbo Ignition\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/0axF9qJyirI\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Phenitia Park.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eWarehouse Rage\u003c/h3\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/VmO_zsf-ccY\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eCinematic race footage: Odessa Keys.\u003c/em\u003e\u003c/p\u003e\n","excerpt":"Have a listen to NullG’s new soundtrack: seven Suno-made instrumental tracks inspired by 90s drum and bass, jungle and EDM.","metaTitle":"A new soundtrack for NullG","metaDescription":"Meet NullG’s new soundtrack: seven Suno-made tracks inspired by 90s electronic music and CoLD SToRAGE, with full music videos to enjoy.","status":"published","featured":false,"publishedAt":"2026-10-22T20:21:53Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"9d8a0e7e68d375fa500374848444d5ff","projectId":"ae4bdb66da6dc1729a263a475e1cdbe3","tags":[{"id":"5af9f6e8d2da4d2b758b298ba0f8057e","name":"Music","slug":"music"},{"id":"10a974c0a281d4fcf3d44f5ef8dae079","name":"NullG","slug":"nullg"}],"createdAt":"2026-10-22T20:19:58Z","updatedAt":"2026-10-22T20:26:56Z"},{"id":"59ab1e54b3a5386e6900ca8678d4e839","siteId":"projectable","slug":"nullg-reinforcement-learning-journey","title":"Teaching NullG to race: the reinforcement-learning journey so far","markdown":"[NullG](https://projectable.me/projects/nullg/) now has a reinforcement-learning pipeline that can train a pilot in the native racing simulation, export its learned controls into the game, and record complete races for review. The latest footage follows that work into weapon combat: preserving a learned driver while training decisions about when and how to use its inventory.\n\nGetting here has involved several changes of direction. Road observations, steering, rewards, exploration, Turbo handling and combat timing have each needed their own experiments. A useful race raises the next question: which part of the system produced that behaviour, and does it hold up on another circuit?\n\nThe [NullG YouTube playlist](https://www.youtube.com/playlist?list=PLaR-_mRLlb6E) follows the progression through recorded driving and combat runs. This post explains the engineering behind those videos, the external projects informing the approach, and the repeated work between longer training sessions.\n\n## A pilot that drives through the game’s controls\n\nNullG’s starting point is its own C++ racing simulation. Python training tools communicate with isolated native workers, each running a race with the engine’s physics, pickups, weapons and race rules. Gymnasium provides the environment interface, and PyTorch with [Stable-Baselines3 PPO](https://stable-baselines3.readthedocs.io/en/master/modules/ppo.html) provides the main training implementation.\n\nA policy is the neural network that turns observations into controls. Its companion during training, the critic, estimates the return expected from a state. PPO collects experience with the current policy, then makes bounded updates to those networks. In NullG, that experience comes from the same native simulation used by gameplay.\n\nThe pilot receives structured observations: its motion and health, connected road geometry ahead, nearby racers, available pickups, visible projectiles, inventory and previous controls. This gives development direct access to questions such as how far the pilot can see into a corner and how it represents lateral movement across the road.\n\nThe outputs correspond to familiar inputs: throttle, analog steering, left and right airbrakes, pitch and weapon commands. Decisions arrive at 30 Hz through the normal player-input path, while the session advances at 60 Hz with the selected source physics cadence. Movement is held between decisions; weapon commands follow the engine’s input timing.\n\nOnce trained, the actor exports into a compact `.ngpolicy` package. A native C++ evaluator reads the observations and produces controls inside NullG. Python remains part of the development workflow, while the exported pilot runs in the engine. Package metadata records the observation contract, model identity and supported configuration, making the training-to-game handoff inspectable.\n\n![](/content/images/2026/10/b1b8220f350c-nullg-rl-full-title-preview.png)\n\n*The latest recorded evaluation race, presented through NullG’s cinematic replay exporter. The driving policy and the replay camera are separate systems.*\n\n## External projects that shaped the approach\n\nRacing RL already has a substantial body of practical work. NullG uses those projects as references for particular engineering decisions, adapting the ideas to an anti-gravity racer and a native engine.\n\n[Sony AI’s Gran Turismo Sophy training account](https://ai.sony/blog/how-to-train-your-race-car) is a reference for building a varied training regimen. Solo driving, interaction with traffic and specialised situations each contribute useful experience. For NullG, that encourages deliberate coverage of circuits, traffic conditions and difficult states, with several kinds of practice available during development.\n\n[TMRL](https://github.com/trackmania-rl/tmrl), the TrackMania reinforcement-learning framework, is useful for its treatment of real-time control, observation histories, analog actions and off-policy learning. It helped motivate a controlled SAC steering comparison alongside PPO. NullG’s native worker supplies a different integration path, while the timing and control questions transfer well.\n\n[Linesight](https://linesight-rl.github.io/linesight/build/html/main_objects.html) provides another TrackMania reference, including its IQN-based approach and explicit reference-line geometry. Its value here is the care given to defining progress and training coverage. [Learn-to-Race](https://github.com/learn-to-race/l2r) adds a useful autonomous-racing environment and baseline perspective: conventional and learned controllers can be compared through a clearly defined task.\n\n[CAPS](https://ai.bu.edu/caps/) informs the smooth-control studies. Action variation and the frequency content of motion provide ways to examine steering quality. That sits alongside a separate question about exploration: the sampled actions used while learning and the deterministic controls used by an exported actor can behave differently.\n\n[NeuralKart](https://github.com/rameshvarun/NeuralKart) supplies a practical reference for searching for useful driving actions and learning from them. NullG adapts that principle using native snapshots from states reached by its own learner. [DAgger](https://proceedings.mlr.press/v15/ross11a.html) gives the broader rationale for collecting corrective examples where the learner actually drives, as successive actions change the states it encounters.\n\nThese references contribute different pieces: curriculum design, control timing, progress geometry, smoothness, baseline comparisons and corrective data collection. NullG combines them with its existing simulation, input and replay systems.\n\n## Giving the driver useful information\n\nEarly work established the loop from observation to action, followed by closer examination of what the network was actually receiving. That led to an audit of observation scaling, feature connectivity, road preview and the steering adapter.\n\nScaling deserves attention because a number can be technically present yet convey very little useful variation after normalisation and clipping. The corrected foundation uses explicit physical scales and identity normalisation for its observation contract. Checks also trace new features through to the network inputs, so a change to the observation layout has a verifiable effect on the model being trained.\n\nRoad guidance then became more explicit. The pilot receives road-forward directions, heading error, longitudinal and lateral velocity, displacement across the road, curvature and available clearance. Connected-road projection gives those measurements a continuous position along the route, including the neighbourhood of section boundaries and qualified junctions.\n\nThat same care applies to progress rewards. Canonical road distance and high-water accounting make forward progress meaningful across the track’s connected geometry. Pickups and boost pads have their own bounded contributions. A reward can then describe a useful driving objective with a clear relationship to what happened on the circuit.\n\nSteering also moved to a native analog contract, with 513 signed positions available through the XL input adapter. A controlled comparison could then ask about the observation design while holding the control protocol and experience budget steady. Another steering study compared PPO and SAC under the same restricted auxiliary controls. Those experiments narrowed individual design choices before the full driver took on more responsibilities.\n\n## The repeated work between longer runs\n\nThe development record across October 15–20 contains an ongoing cycle of collecting races, reviewing telemetry, making a focused adjustment, and testing the resulting policy again. This has occupied a substantial part of the RL work alongside the training itself.\n\nSeveral strategies needed revisiting. Increasing a steering-change penalty could make sampled inputs look quieter, so review expanded to the policy’s mean output and the ship’s actual heading and lateral motion. Longer road preview and different brake commitment were tested separately. Exploration settings were examined alongside deterministic native inference, giving both the training behaviour and the exported driver a direct comparison.\n\nTurbo made preservation especially important. A steering change affects the trajectory from which Turbo is used, even when the weapon-control weights stay identical. Naturally earned Turbo states therefore became part of the review panels and corrective training anchors. Useful speed, health, recovery and the rest of the race all contribute to judging the manoeuvre.\n\nThe corrective-driving work took another route. Native snapshots preserve a state reached by the learner, then development tooling tries small steering interventions from that same state. Accepted examples become training labels, with unchanged actions retained as anchors. Focused work on Gare d’Europa and Odessa Keys added longer follow-through checks and preservation samples from the other circuits.\n\nEach revision brought additional work around the optimiser: reconstructing states, verifying the replayed baseline, collecting examples, fitting a bounded update, exporting the actor and running complete races. Longer safety follow-through was added to examine the consequences beyond a promising short steering branch.\n\nThe time involved can be described concretely from the run receipts. Three completed combat jobs alone—the eight-hour run on October 18, the four-hour run on October 19, and the latest four-hour continuation—account for **more than 16 hours of training wall time**. Short comparison arms, earlier driving work, evaluation races, corrective searches and media production add to that. Those three jobs branch from selected earlier checkpoints, so their durations describe compute spent across experiments rather than sixteen uninterrupted hours on one policy.\n\nThe records establish those training durations. Reading references, inspecting footage and revising the tools also take development time; that human work has no consolidated hour counter in the run receipts. The visible pattern is a sequence of experiments and adjustments over several days, with longer jobs following more focused checks.\n\n## From driving practice to racing and combat\n\nThe driving foundation gave the pilot room to practise navigation, traffic, pickups and earned Turbo. Combat then introduced a different timing problem: choosing an action now can affect another racer several seconds later, and that exchange can change the final placing.\n\nThe approach evolved toward training the weapon outputs and critic while freezing the actor’s driving features and movement outputs. That lets the combat experiment build on an existing driver. Weight checks verify the protected parameters, while complete races show how the new combat choices affect the resulting trajectories.\n\nThe objective also became explicitly placement-first. Race outcome has the dominant reward, with bounded auxiliary credit for effective combat and useful pickups or pads. Shot count by itself receives no bonus. The critic and credit-assignment settings then become important parts of learning when a weapon opportunity is worth taking.\n\nExperiments examined longer credit assignment, combat reward weighting and critic calibration. Calibration fits the value estimate to returns collected from the retained policy, while preserving the actor. Opportunity starts reconstruct naturally reached situations with offensive inventory and nearby opponents, giving the learner more practice around decisions that may be relatively sparse in ordinary races.\n\nThe latest configuration combines ordinary races with ten-second combat opportunity windows. These windows start from verified native states and collect fresh actions from the current policy. Their training cutoff is handled separately from a race finish or elimination, preserving the meaning of the full-race objective.\n\nTraining opposition has expanded too: original XL opponents, a retained learned challenger, and mixed fields provide different contexts. The active development rotation is Talon’s Reach, Sagarmatha, Gare d’Europa and Odessa Keys.\n\n## What the latest four-hour continuation did\n\nThe latest completed continuation began at the earlier combat run’s hour-two checkpoint. Four CPU workers added **6,639,616 decisions**, moving the checkpoint counter from **32,651,264 to 39,290,880**. That number belongs to this checkpoint lineage; it is separate from the experience collected across all the other experiments.\n\nThe run preserved the driving parameters and continued learning weapon outputs and the critic. Its recorded settings retained a placement-first objective, gamma 1.0, GAE lambda 0.95, combat reward scale 1.0 and 300-decision opportunity windows. In practical terms, the experiment concentrated on combat choices while retaining the driver’s learned control function.\n\nHourly checkpoints made it possible to compare behaviour during the continuation. The subsequent review covered six policies, with 48 standard evaluation races per policy and additional cross-play and confirmation panels. [RLiable’s evaluation guidance](https://research.google/blog/rliable-towards-reliable-evaluation-reporting-in-reinforcement-learning/) informs this emphasis on paired conditions, multiple seeds and the distribution of results.\n\nThe final checkpoint won 18 of its 48 standard-panel races and finished 31. In the additional 24-race confirmation panel, it won 9 and finished 15. Different intermediate checkpoints had different strengths, so the review retains those trade-offs for the next decision. A cinematic race shows the driving and combat in detail; the larger panels establish the context for interpreting it.\n\n![](/content/images/2026/10/b1e12503f058-nullg-rl-full-preview.png)\n\n*Recorded Gare d’Europa gameplay from the final checkpoint. The cinematic director follows racing interest and weapon exchanges across the field, including opponents as well as the learned racer.*\n\n## Watch the latest races\n\nThe videos make the development easier to follow alongside the technical account. Earlier driving updates show the broader training progression; the latest full race and Short focus on racing with weapons. The learned pilot occupies slot 0 in the latest full race, while automatic cinematic cameras also cover other racers and exchanges.\n\n### Learning to fight: the complete Gare d’Europa race\n\nThis complete three-lap race comes from the final checkpoint of the latest four-hour continuation. It runs for approximately three minutes, with native race audio and cinematic presentation.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/__kpDHYbsCg\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n[Watch the full Gare d’Europa race on YouTube](https://youtu.be/__kpDHYbsCg).\n\n### Learning to race and fight: the Short\n\nThe latest Short offers a compact look at the racing and combat work.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/lTbLUj1QKt4\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n[Watch the racing and combat on YouTube](https://youtu.be/lTbLUj1QKt4).\n\n### Earlier training footage: update 3\n\nThis earlier upload is titled “NullG - 8 hour AI driving training 3 (reinforcement learning)” and forms part of the playlist’s training record.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/fkUTwW2p_8k\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n[Watch training update 3 on YouTube](https://youtu.be/fkUTwW2p_8k).\n\n### Earlier training footage: update 2\n\nThe preceding “NullG - 8 hour AI driving training 2 (reinforcement learning)” upload provides another point in that recorded progression.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/9ddeaoyyIY4\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n[Watch training update 2 on YouTube](https://youtu.be/9ddeaoyyIY4).\n\n## Following the next experiments\n\nThe work now has a native training loop, portable in-game policies, corrective driving tools, combat-focused continuations and recorded evaluations to guide further changes. The next direction is to build on the checkpoint comparisons: keep race outcome central, improve the usefulness of weapon decisions, and preserve driving quality across the four-circuit rotation.\n\nThe [reinforcement-learning project page](https://projectable.me/projects/nullg/reinforcement-learning/) describes the engine integration. The [YouTube playlist](https://www.youtube.com/playlist?list=PLaR-_mRLlb6E) follows the training through the races themselves, with the full recordings providing context for each shorter highlight.","html":"\u003cp\u003e\u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG\u003c/a\u003e now has a reinforcement-learning pipeline that can train a pilot in the native racing simulation, export its learned controls into the game, and record complete races for review. The latest footage follows that work into weapon combat: preserving a learned driver while training decisions about when and how to use its inventory.\u003c/p\u003e\n\u003cp\u003eGetting here has involved several changes of direction. Road observations, steering, rewards, exploration, Turbo handling and combat timing have each needed their own experiments. A useful race raises the next question: which part of the system produced that behaviour, and does it hold up on another circuit?\u003c/p\u003e\n\u003cp\u003eThe \u003ca href=\"https://www.youtube.com/playlist?list=PLaR-_mRLlb6E\" rel=\"nofollow\"\u003eNullG YouTube playlist\u003c/a\u003e follows the progression through recorded driving and combat runs. This post explains the engineering behind those videos, the external projects informing the approach, and the repeated work between longer training sessions.\u003c/p\u003e\n\u003ch2\u003eA pilot that drives through the game’s controls\u003c/h2\u003e\n\u003cp\u003eNullG’s starting point is its own C++ racing simulation. Python training tools communicate with isolated native workers, each running a race with the engine’s physics, pickups, weapons and race rules. Gymnasium provides the environment interface, and PyTorch with \u003ca href=\"https://stable-baselines3.readthedocs.io/en/master/modules/ppo.html\" rel=\"nofollow\"\u003eStable-Baselines3 PPO\u003c/a\u003e provides the main training implementation.\u003c/p\u003e\n\u003cp\u003eA policy is the neural network that turns observations into controls. Its companion during training, the critic, estimates the return expected from a state. PPO collects experience with the current policy, then makes bounded updates to those networks. In NullG, that experience comes from the same native simulation used by gameplay.\u003c/p\u003e\n\u003cp\u003eThe pilot receives structured observations: its motion and health, connected road geometry ahead, nearby racers, available pickups, visible projectiles, inventory and previous controls. This gives development direct access to questions such as how far the pilot can see into a corner and how it represents lateral movement across the road.\u003c/p\u003e\n\u003cp\u003eThe outputs correspond to familiar inputs: throttle, analog steering, left and right airbrakes, pitch and weapon commands. Decisions arrive at 30 Hz through the normal player-input path, while the session advances at 60 Hz with the selected source physics cadence. Movement is held between decisions; weapon commands follow the engine’s input timing.\u003c/p\u003e\n\u003cp\u003eOnce trained, the actor exports into a compact \u003ccode\u003e.ngpolicy\u003c/code\u003e package. A native C++ evaluator reads the observations and produces controls inside NullG. Python remains part of the development workflow, while the exported pilot runs in the engine. Package metadata records the observation contract, model identity and supported configuration, making the training-to-game handoff inspectable.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/b1b8220f350c-nullg-rl-full-title-preview.png\" alt=\"\"\u003e\n\u003cp\u003e\u003cem\u003eThe latest recorded evaluation race, presented through NullG’s cinematic replay exporter. The driving policy and the replay camera are separate systems.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eExternal projects that shaped the approach\u003c/h2\u003e\n\u003cp\u003eRacing RL already has a substantial body of practical work. NullG uses those projects as references for particular engineering decisions, adapting the ideas to an anti-gravity racer and a native engine.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://ai.sony/blog/how-to-train-your-race-car\" rel=\"nofollow\"\u003eSony AI’s Gran Turismo Sophy training account\u003c/a\u003e is a reference for building a varied training regimen. Solo driving, interaction with traffic and specialised situations each contribute useful experience. For NullG, that encourages deliberate coverage of circuits, traffic conditions and difficult states, with several kinds of practice available during development.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://github.com/trackmania-rl/tmrl\" rel=\"nofollow\"\u003eTMRL\u003c/a\u003e, the TrackMania reinforcement-learning framework, is useful for its treatment of real-time control, observation histories, analog actions and off-policy learning. It helped motivate a controlled SAC steering comparison alongside PPO. NullG’s native worker supplies a different integration path, while the timing and control questions transfer well.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://linesight-rl.github.io/linesight/build/html/main_objects.html\" rel=\"nofollow\"\u003eLinesight\u003c/a\u003e provides another TrackMania reference, including its IQN-based approach and explicit reference-line geometry. Its value here is the care given to defining progress and training coverage. \u003ca href=\"https://github.com/learn-to-race/l2r\" rel=\"nofollow\"\u003eLearn-to-Race\u003c/a\u003e adds a useful autonomous-racing environment and baseline perspective: conventional and learned controllers can be compared through a clearly defined task.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://ai.bu.edu/caps/\" rel=\"nofollow\"\u003eCAPS\u003c/a\u003e informs the smooth-control studies. Action variation and the frequency content of motion provide ways to examine steering quality. That sits alongside a separate question about exploration: the sampled actions used while learning and the deterministic controls used by an exported actor can behave differently.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://github.com/rameshvarun/NeuralKart\" rel=\"nofollow\"\u003eNeuralKart\u003c/a\u003e supplies a practical reference for searching for useful driving actions and learning from them. NullG adapts that principle using native snapshots from states reached by its own learner. \u003ca href=\"https://proceedings.mlr.press/v15/ross11a.html\" rel=\"nofollow\"\u003eDAgger\u003c/a\u003e gives the broader rationale for collecting corrective examples where the learner actually drives, as successive actions change the states it encounters.\u003c/p\u003e\n\u003cp\u003eThese references contribute different pieces: curriculum design, control timing, progress geometry, smoothness, baseline comparisons and corrective data collection. NullG combines them with its existing simulation, input and replay systems.\u003c/p\u003e\n\u003ch2\u003eGiving the driver useful information\u003c/h2\u003e\n\u003cp\u003eEarly work established the loop from observation to action, followed by closer examination of what the network was actually receiving. That led to an audit of observation scaling, feature connectivity, road preview and the steering adapter.\u003c/p\u003e\n\u003cp\u003eScaling deserves attention because a number can be technically present yet convey very little useful variation after normalisation and clipping. The corrected foundation uses explicit physical scales and identity normalisation for its observation contract. Checks also trace new features through to the network inputs, so a change to the observation layout has a verifiable effect on the model being trained.\u003c/p\u003e\n\u003cp\u003eRoad guidance then became more explicit. The pilot receives road-forward directions, heading error, longitudinal and lateral velocity, displacement across the road, curvature and available clearance. Connected-road projection gives those measurements a continuous position along the route, including the neighbourhood of section boundaries and qualified junctions.\u003c/p\u003e\n\u003cp\u003eThat same care applies to progress rewards. Canonical road distance and high-water accounting make forward progress meaningful across the track’s connected geometry. Pickups and boost pads have their own bounded contributions. A reward can then describe a useful driving objective with a clear relationship to what happened on the circuit.\u003c/p\u003e\n\u003cp\u003eSteering also moved to a native analog contract, with 513 signed positions available through the XL input adapter. A controlled comparison could then ask about the observation design while holding the control protocol and experience budget steady. Another steering study compared PPO and SAC under the same restricted auxiliary controls. Those experiments narrowed individual design choices before the full driver took on more responsibilities.\u003c/p\u003e\n\u003ch2\u003eThe repeated work between longer runs\u003c/h2\u003e\n\u003cp\u003eThe development record across October 15–20 contains an ongoing cycle of collecting races, reviewing telemetry, making a focused adjustment, and testing the resulting policy again. This has occupied a substantial part of the RL work alongside the training itself.\u003c/p\u003e\n\u003cp\u003eSeveral strategies needed revisiting. Increasing a steering-change penalty could make sampled inputs look quieter, so review expanded to the policy’s mean output and the ship’s actual heading and lateral motion. Longer road preview and different brake commitment were tested separately. Exploration settings were examined alongside deterministic native inference, giving both the training behaviour and the exported driver a direct comparison.\u003c/p\u003e\n\u003cp\u003eTurbo made preservation especially important. A steering change affects the trajectory from which Turbo is used, even when the weapon-control weights stay identical. Naturally earned Turbo states therefore became part of the review panels and corrective training anchors. Useful speed, health, recovery and the rest of the race all contribute to judging the manoeuvre.\u003c/p\u003e\n\u003cp\u003eThe corrective-driving work took another route. Native snapshots preserve a state reached by the learner, then development tooling tries small steering interventions from that same state. Accepted examples become training labels, with unchanged actions retained as anchors. Focused work on Gare d’Europa and Odessa Keys added longer follow-through checks and preservation samples from the other circuits.\u003c/p\u003e\n\u003cp\u003eEach revision brought additional work around the optimiser: reconstructing states, verifying the replayed baseline, collecting examples, fitting a bounded update, exporting the actor and running complete races. Longer safety follow-through was added to examine the consequences beyond a promising short steering branch.\u003c/p\u003e\n\u003cp\u003eThe time involved can be described concretely from the run receipts. Three completed combat jobs alone—the eight-hour run on October 18, the four-hour run on October 19, and the latest four-hour continuation—account for \u003cstrong\u003emore than 16 hours of training wall time\u003c/strong\u003e. Short comparison arms, earlier driving work, evaluation races, corrective searches and media production add to that. Those three jobs branch from selected earlier checkpoints, so their durations describe compute spent across experiments rather than sixteen uninterrupted hours on one policy.\u003c/p\u003e\n\u003cp\u003eThe records establish those training durations. Reading references, inspecting footage and revising the tools also take development time; that human work has no consolidated hour counter in the run receipts. The visible pattern is a sequence of experiments and adjustments over several days, with longer jobs following more focused checks.\u003c/p\u003e\n\u003ch2\u003eFrom driving practice to racing and combat\u003c/h2\u003e\n\u003cp\u003eThe driving foundation gave the pilot room to practise navigation, traffic, pickups and earned Turbo. Combat then introduced a different timing problem: choosing an action now can affect another racer several seconds later, and that exchange can change the final placing.\u003c/p\u003e\n\u003cp\u003eThe approach evolved toward training the weapon outputs and critic while freezing the actor’s driving features and movement outputs. That lets the combat experiment build on an existing driver. Weight checks verify the protected parameters, while complete races show how the new combat choices affect the resulting trajectories.\u003c/p\u003e\n\u003cp\u003eThe objective also became explicitly placement-first. Race outcome has the dominant reward, with bounded auxiliary credit for effective combat and useful pickups or pads. Shot count by itself receives no bonus. The critic and credit-assignment settings then become important parts of learning when a weapon opportunity is worth taking.\u003c/p\u003e\n\u003cp\u003eExperiments examined longer credit assignment, combat reward weighting and critic calibration. Calibration fits the value estimate to returns collected from the retained policy, while preserving the actor. Opportunity starts reconstruct naturally reached situations with offensive inventory and nearby opponents, giving the learner more practice around decisions that may be relatively sparse in ordinary races.\u003c/p\u003e\n\u003cp\u003eThe latest configuration combines ordinary races with ten-second combat opportunity windows. These windows start from verified native states and collect fresh actions from the current policy. Their training cutoff is handled separately from a race finish or elimination, preserving the meaning of the full-race objective.\u003c/p\u003e\n\u003cp\u003eTraining opposition has expanded too: original XL opponents, a retained learned challenger, and mixed fields provide different contexts. The active development rotation is Talon’s Reach, Sagarmatha, Gare d’Europa and Odessa Keys.\u003c/p\u003e\n\u003ch2\u003eWhat the latest four-hour continuation did\u003c/h2\u003e\n\u003cp\u003eThe latest completed continuation began at the earlier combat run’s hour-two checkpoint. Four CPU workers added \u003cstrong\u003e6,639,616 decisions\u003c/strong\u003e, moving the checkpoint counter from \u003cstrong\u003e32,651,264 to 39,290,880\u003c/strong\u003e. That number belongs to this checkpoint lineage; it is separate from the experience collected across all the other experiments.\u003c/p\u003e\n\u003cp\u003eThe run preserved the driving parameters and continued learning weapon outputs and the critic. Its recorded settings retained a placement-first objective, gamma 1.0, GAE lambda 0.95, combat reward scale 1.0 and 300-decision opportunity windows. In practical terms, the experiment concentrated on combat choices while retaining the driver’s learned control function.\u003c/p\u003e\n\u003cp\u003eHourly checkpoints made it possible to compare behaviour during the continuation. The subsequent review covered six policies, with 48 standard evaluation races per policy and additional cross-play and confirmation panels. \u003ca href=\"https://research.google/blog/rliable-towards-reliable-evaluation-reporting-in-reinforcement-learning/\" rel=\"nofollow\"\u003eRLiable’s evaluation guidance\u003c/a\u003e informs this emphasis on paired conditions, multiple seeds and the distribution of results.\u003c/p\u003e\n\u003cp\u003eThe final checkpoint won 18 of its 48 standard-panel races and finished 31. In the additional 24-race confirmation panel, it won 9 and finished 15. Different intermediate checkpoints had different strengths, so the review retains those trade-offs for the next decision. A cinematic race shows the driving and combat in detail; the larger panels establish the context for interpreting it.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/b1e12503f058-nullg-rl-full-preview.png\" alt=\"\"\u003e\n\u003cp\u003e\u003cem\u003eRecorded Gare d’Europa gameplay from the final checkpoint. The cinematic director follows racing interest and weapon exchanges across the field, including opponents as well as the learned racer.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eWatch the latest races\u003c/h2\u003e\n\u003cp\u003eThe videos make the development easier to follow alongside the technical account. Earlier driving updates show the broader training progression; the latest full race and Short focus on racing with weapons. The learned pilot occupies slot 0 in the latest full race, while automatic cinematic cameras also cover other racers and exchanges.\u003c/p\u003e\n\u003ch3\u003eLearning to fight: the complete Gare d’Europa race\u003c/h3\u003e\n\u003cp\u003eThis complete three-lap race comes from the final checkpoint of the latest four-hour continuation. It runs for approximately three minutes, with native race audio and cinematic presentation.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/__kpDHYbsCg\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003ca href=\"https://youtu.be/__kpDHYbsCg\" rel=\"nofollow\"\u003eWatch the full Gare d’Europa race on YouTube\u003c/a\u003e.\u003c/p\u003e\n\u003ch3\u003eLearning to race and fight: the Short\u003c/h3\u003e\n\u003cp\u003eThe latest Short offers a compact look at the racing and combat work.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/lTbLUj1QKt4\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003ca href=\"https://youtu.be/lTbLUj1QKt4\" rel=\"nofollow\"\u003eWatch the racing and combat on YouTube\u003c/a\u003e.\u003c/p\u003e\n\u003ch3\u003eEarlier training footage: update 3\u003c/h3\u003e\n\u003cp\u003eThis earlier upload is titled “NullG - 8 hour AI driving training 3 (reinforcement learning)” and forms part of the playlist’s training record.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/fkUTwW2p_8k\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003ca href=\"https://youtu.be/fkUTwW2p_8k\" rel=\"nofollow\"\u003eWatch training update 3 on YouTube\u003c/a\u003e.\u003c/p\u003e\n\u003ch3\u003eEarlier training footage: update 2\u003c/h3\u003e\n\u003cp\u003eThe preceding “NullG - 8 hour AI driving training 2 (reinforcement learning)” upload provides another point in that recorded progression.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/9ddeaoyyIY4\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003ca href=\"https://youtu.be/9ddeaoyyIY4\" rel=\"nofollow\"\u003eWatch training update 2 on YouTube\u003c/a\u003e.\u003c/p\u003e\n\u003ch2\u003eFollowing the next experiments\u003c/h2\u003e\n\u003cp\u003eThe work now has a native training loop, portable in-game policies, corrective driving tools, combat-focused continuations and recorded evaluations to guide further changes. The next direction is to build on the checkpoint comparisons: keep race outcome central, improve the usefulness of weapon decisions, and preserve driving quality across the four-circuit rotation.\u003c/p\u003e\n\u003cp\u003eThe \u003ca href=\"https://projectable.me/projects/nullg/reinforcement-learning/\" rel=\"nofollow\"\u003ereinforcement-learning project page\u003c/a\u003e describes the engine integration. The \u003ca href=\"https://www.youtube.com/playlist?list=PLaR-_mRLlb6E\" rel=\"nofollow\"\u003eYouTube playlist\u003c/a\u003e follows the training through the races themselves, with the full recordings providing context for each shorter highlight.\u003c/p\u003e\n","excerpt":"From native driving controls and road guidance to Turbo preservation and weapon combat: the experiments, references and repeated adjustments behind NullG’s learning pilots.","metaTitle":"NullG’s reinforcement-learning journey: driving, Turbo and combat","metaDescription":"Explore NullG’s native racing RL pipeline, external research references, strategy revisions and latest training videos, from analog steering to combat.","status":"published","featured":false,"publishedAt":"2026-10-20T13:21:42Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"2de6b37e96d56188d9b74019123198cd","projectId":"ae4bdb66da6dc1729a263a475e1cdbe3","tags":[{"id":"10a974c0a281d4fcf3d44f5ef8dae079","name":"NullG","slug":"nullg"},{"id":"1c9e8a1339a8f8b797767e91a9673016","name":"Game","slug":"game"},{"id":"60103f47eb65b46c6138b9d6fc7b17d0","name":"Machine Learning","slug":"machine-learning"}],"createdAt":"2026-10-20T13:10:40Z","updatedAt":"2026-10-20T15:07:55Z"},{"id":"7b367212a739706ba4e3e9b999bec15d","siteId":"projectable","slug":"introducing-nullg-anti-gravity-racing-and-ai","title":"Introducing NullG: anti-gravity racing and AI-assisted development","markdown":"[NullG](https://projectable.me/projects/nullg/) is an anti-gravity racing engine built in C++20 with Vulkan for rendering. It combines legacy game asset imports with racing physics, weapons, multiplayer, and tools for testing and capturing gameplay.\n\nThe longer-term goal is a motion simulator pod, with acceleration, braking, cornering, and impacts feeding into physical movement. Positional audio and multiple displays are part of that plan too. Earlier projects such as the [Car-cade](https://projectable.me/my-dirt-rally-tabletop-car-cade-game/) and [motion rig](https://projectable.me/sim-racing-motion-rig-v2-updates/) explored racing through custom hardware. NullG takes that work into the game engine.\n\nDevelopment started in January 2026 with AI-assisted planning and implementation. Over the following months, the work expanded into asset loading, rendering, physics, and automated tests. October added more direct access for agents to inspect the engine, run controlled scenarios, and review the results.\n\n![Original NullG ship artwork](/content/images/2026/10/583bdbe059a6-ship.png)\n\n*Original NullG ship artwork.*\n\n## January: plans and the first renderer\n\nThe first git commit, dated January 22, 2026, contains the project outline, research files, architecture notes, and initial application code. The plans already covered asset conversion, input, audio, networking, computer-controlled opponents, and motion output. The application had a window, Vulkan setup, input handling, configuration, and logging.\n\nThe architecture put the race simulation at the center. Graphics and sound would use the state produced by the simulation, with motion hardware eventually reading the same data. This kept the simulator goal part of the engine design from the beginning.\n\nEarly development used Cursor, with project commands for planning, preparation, and review. The preparation step directed the agent to read reference material and trace the relevant code before making changes. Tasks were broken into smaller pieces, with examples and written decisions available for the next session.\n\n![G_jZiq3WQAAg2Cz.jpeg](/content/images/2026/10/037075fac8ed-G_jZiq3WQAAg2Cz.jpeg)\n\nJanuary 23 brought asset conversion tooling and the first renderer pass. Menus, model loading, asset management, and track parsing followed over the next few days. Existing WipEout research and community projects provided references for loading the ships, tracks, textures, and scenery.\n\n## February: loading the original assets\n\nBy February 8, the project had a direct PlayStation asset-loading checkpoint. Work during this period covered track rendering, scene objects, sprites, and texture transparency.\n\nLoading the geometry, textures, and scenery together made it possible to check the imported data in the renderer. File-format notes and agent rules developed alongside the loaders, giving later tasks a record of how the assets were organized and used.\n\n## March: handling, race flow, and weapons\n\nMarch focused more on how ships move around a track: hover height, coasting, weight, jumps, boost, recovery, and tunnel-wall collisions. Scene placement, sky textures, pickups, and boost pads also received attention, along with the camera moving through the track behind the menus.\n\nReal-time lighting, shadows, and GPU particles added to the presentation. By March 6, the code included race-session flow and a weapon system, handling the transition into a race and back to the menus.\n\nThese changes were developed as individual tasks using the relevant code and reference material. Code review and updated implementation notes kept the decisions available as the systems grew.\n\n## May: rendering and automated tests\n\nMay brought a larger pass through the engine internals. The original engine plan was split into focused documents with a shared status index, making individual areas easier to track.\n\nRenderer work included dynamic rendering, geometry access, material organization, and a move to Slang for runtime shaders. Lighting work covered animated scene lights, illuminated pickups, shields, and the shadows received by ships from surrounding scenery.\n\nAn entity/component scene system and render snapshots were added during the same period. A snapshot holds the scene data needed for a frame, including ship and object positions, giving the renderer a consistent set of data to draw.\n\nOn May 28, a headless test suite added checks for the job system, memory, and asset handling. Agents could run these tests after editing the code and inspect the results directly. This became the starting point for the more extensive inspection and gameplay tools added later.\n\n## June: collision, effects, and XL\n\nJune added Jolt collision integration and further GPU effects work. Research notes covered established techniques for effects, shadows, and physics, providing examples for subsequent implementation tasks.\n\nWipEout XL became the default asset source in a June 9 checkpoint. Source descriptors drove the menus, allowing the selected game's data to determine the available choices. Other changes covered waterfall texture scrolling and the starting-grid camera.\n\nBy this stage, the codebase covered rendering, asset loading, handling, race flow, weapons, collision, and tests. The next concentrated stretch of commits begins on October 9, with the existing agent guidance brought into Codex, engine-backed inspection tools, and a fixed 60 Hz racing setup. Multiplayer followed on October 10, then the XL reconstruction checkpoint on October 11.\n\n## October: giving agents access to the engine\n\nOctober also marked a decision to use OpenAI Codex exclusively for NullG's coding-agent work. The existing planning commands, project rules, and reference notes were carried over from Cursor, keeping the earlier work available as context. Development then focused more heavily on giving Codex access to the tools needed to inspect, test, and review changes.\n\nRecently watched YouTube videos about Blender MCP helped shape that approach. They showed how an agent could work directly with a scene and inspect the result. Those concepts were brought into NullG's model-authoring process and reinforced the value of connecting agents to the engine's own inspection tools.\n\nOctober's tooling work gave agents more ways to check their changes. [MCP, the Model Context Protocol](https://modelcontextprotocol.io/introduction), connects AI applications to external tools and data. NullG uses it to expose native inspection windows, track validators, model viewers, and controlled race sessions.\n\nAn agent can load a track, choose a camera, inspect geometry, and capture a frame. Model inspection provides access to materials, bounds, and different viewing angles. Test harnesses provide repeatable scenarios with known inputs.\n\n![NullG Track Validator showing track geometry, texture inspection, and timeline controls](/content/images/2026/10/nullg/intro-track-inspection.png)\n\n*The track inspector, with the track view, texture panel, and playback controls.*\n\nThe weapon harness opens a paused production race scene. An agent can deploy a weapon and advance the scene in controlled steps, inspecting a projectile, impact, or Quake effect at a specific point. Simulation state and rendered output can then be reviewed together.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/pbUzB3j74p8\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*Reactive engine plumes viewed through the MCP harness.*\n\nPhysics tests compare movement with expected values, track checks inspect road geometry, and screenshots show the rendered result. Replaying the same inputs checks consistency across changes. Together, these tools let agents run a larger part of the validation process themselves, with reports and captures available for human review.\n\n![Ships racing with engine trails and a weapon effect in NullG](/content/images/2026/10/nullg/nullg-ships-weapon-action.png)\n\n### Working through a change\n\nA weapon-effect investigation starts with a specific scene and setup. The agent deploys the weapon, steps through the action, checks the state, and follows the code responsible for the effect. After an edit, it builds the affected code and repeats the scenario.\n\nSaved reports retain the inputs and settings. That allows the case to be repeated in another session or compared across versions. Visual review and gameplay testing provide additional checks on appearance and handling.\n\n![Diagram showing an agent connected through MCP to native inspection and headless validation tools](/content/images/2026/10/nullg/intro-agent-workflow.png)\n\n*The agent can run a test, open an inspector, and save the results for review.*\n\n## Reverse engineering WipEout XL\n\nThe October research also examined WipEout XL / 2097 behavior in more detail, including steering, weapon updates, starting positions, and animated objects.\n\n[REA, Reverse Engineer Anything](https://github.com/morluto/rea), formed part of the research setup. It connects agents with tools for investigating applications and binaries. The PlayStation XL executable decompilation used a PSX-aware Ghidra workflow, with scripts retaining instructions, references, annotations, and findings.\n\nThe starting grid provides a useful example. Following the original track numbering and route links established where ships should start. Those findings became checks against the imported track, with the inspector showing the selected position on the road.\n\n![An overhead track inspection with a highlighted grid marker](/content/images/2026/10/nullg/intro-grid-placement.png)\n\n*Checking a grid position against the track geometry.*\n\nThe research covers both USA and European editions. Each finding retains the edition and executable details so timing values and behavior can be traced to their source.\n\nRecovered calculations and state changes are implemented as small C++ reference models, tested, and then connected to NullG. The records distinguish behavior derived from the original executable, adaptations for NullG, and observations from running the engine.\n\nAgent access to analysis tools makes this kind of detailed investigation more accessible to individual developers. Existing products can be studied one behavior at a time, with findings carried into tests and new implementations. For NullG, that supports continued refinement of racing behavior alongside modern rendering, controls, and multiplayer. Imported content comes from user-supplied game files, while original NullG assets are developed separately.\n\n## A wider wave of software reconstruction\n\nREA's rapid rise in attention during early October put agent-assisted reverse engineering in front of a much larger audience. [Coverage of its growth](https://www.onfirstlaunch.com/blog/reverse-engineer-anything-rea-explained) described a surge in interest around demonstrations of agents inspecting applications and reconstructing their behavior.\n\nAt the same time, “vibe-coded” alternatives to familiar creative and office software were attracting attention. [ArtCraft's projects](https://github.com/storytold) include photo, video, and PDF tools, alongside [WordCraft](https://github.com/storytold/wordcraft), [GridCraft](https://github.com/storytold/gridcraft), and [DeckCraft](https://github.com/storytold/deckcraft), which target Word, Excel, and PowerPoint-style workflows. [October coverage](https://www.techradar.com/computing/office-suites/developer-who-vibe-coded-free-adobe-clones-reveals-apps-to-take-down-microsoft-office-next-and-says-there-is-no-reason-to-save-broken-software) connected these AI-assisted projects with the growing interest in alternatives to Adobe and Microsoft applications.\n\nThese are separate projects exploring related possibilities: using agents to understand software, implement familiar workflows, and build tools around specific needs. NullG applies that approach to racing. Executable research informs individual behaviors, while tests, inspectors, and recorded scenarios provide ways to check the implementation. The same emphasis on repeatable validation carries through asset production and gameplay development.\n\n## Blender and model production\n\nThe Blender MCP concepts from those videos were applied to a specific job: preparing original menu models for the engine. Scene access let the agent inspect parts, make scoped edits, export a candidate, and produce comparison views within the same task.\n\nThe benefit was practical. Model changes could be checked from several angles in Blender and then reviewed at their actual size in NullG's menus. Saved revisions and repeatable exports made the process easier to iterate, while geometry checks and visual review stayed part of the finishing pass.\n\nBlender MCP provides agent access to scene inspection, geometry and material editing, and visual review. The menu-model process starts with visual references and generated models where useful, followed by a Blender finishing pass.\n\nThat pass covers shape, individual parts, surface cleanup, materials, and polygon count. Models are reviewed from several angles, exported, and checked in NullG's model viewer and menus at their intended display size.\n\n![Four Blender review renders of the original monitor-shaped Graphics menu model](/content/images/2026/10/nullg/intro-blender-graphics-model.png)\n\n*Front, side, rear, and angled review renders of the Graphics menu model.*\n\nThe authoring files, exports, and comparison images remain available through each revision. Connecting these tools lets an agent work through research, code, tests, and visual checks within the same development session.\n\n## Race capture and social media\n\nNullG includes race recording, replay playback, cinematic cameras, and video export. Recorded races can be rendered with ship-focused shots, action cuts, titles, and audio. Landscape, portrait, and square formats support both longer videos and shorter social clips.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/tREm0SM8zmg\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*The XL cinematic race intro.*\n\nScreenshot tools capture menus, starting grids, tracks, and weapons in action. Agents can help select and organize those captures, with settings and source details retained alongside the images. The same recorded race can support a gameplay review and provide footage for a project update.\n\n## Reinforcement learning for the pilots\n\nNullG also includes reinforcement-learning experiments for computer-controlled pilots. A trained model selects ship controls during a race. Development agents help build the training tools, run checks, and prepare the results for review.\n\nThe pilot receives information about speed and direction, the road ahead, nearby ships, and pickups. It selects throttle, steering, air brakes, pitch, and weapon commands, which pass through the racing simulation as player-style inputs.\n\nTraining uses Python, Gymnasium, Stable-Baselines3, and PyTorch, with PPO as the learning algorithm. Native workers run NullG's race simulation, allowing several driving attempts to run in parallel while the training process updates the model.\n\n![](/content/images/2026/10/nullg/intro-rl-city-race.png)\n\n*A frame from an RL evaluation replay on Gare d’Europa, rendered through NullG's replay tools.*\n\n### Training and evaluation\n\nA short run checks the training setup and controls. Longer runs save checkpoints for evaluation on a chosen set of tracks and race configurations. Saved settings and seeds allow those checks to be repeated.\n\nEvaluation measures completion, time, damage, and steering smoothness. Traffic and combat scenarios also examine race position, pickup use, and weapon timing. Training checkpoints and policies approved for gameplay have separate evaluation steps.\n\nReplays provide another view of the results. Chase cameras show ship movement through a corner, while overhead views show racer positions around the circuit. These views can be compared with the recorded controls and race metrics.\n\n![](/content/images/2026/10/nullg/intro-rl-overhead.png)\n\n*An overhead replay view from the RL development work, with colored racer markers.*\n\nThe exported pilot is a compact model evaluated by native C++ code. Python handles training, and the game runs the exported policy. Retained checkpoints allow different training runs to be compared using the same evaluation setup.\n\nAgents can run the checks, read reports, and prepare replay captures. This supports review of individual driving behaviors and produces footage for documenting the experiments.\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/fkUTwW2p_8k\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n*An AI-driving training clip from the NullG playlist.*\n\nMore detail is available on the [MCP tooling page](https://projectable.me/projects/nullg/mcp-tooling/) and [reinforcement-learning page](https://projectable.me/projects/nullg/reinforcement-learning/).\n\n## Publishing to projectable.me\n\nThe same agent-assisted workflow extends to publishing on projectable.me. Agents can prepare illustrated posts, update project pages, and organize development updates under the NullG project.\n\nThe process starts with the git history, project documentation, and relevant captures. An agent prepares the article, adds screenshots and video embeds, and saves a draft on projectable.me. The post can then be reviewed and edited on the site before publication.\n\nThis post follows that process, using repository history for the timeline, project documentation for the workflow details, and existing engine captures for the screenshots. Keeping these updates attached to the NullG project makes the development history easy to follow alongside the project pages.\n\n## Current state and next steps\n\nNullG now includes racing, weapons, computer-controlled opponents, local split-screen, LAN and private online rooms, and replay tools. Inspection and validation tools support development across those systems.\n\n![Two-player vertical split-screen racing with WipEout XL content in NullG](/content/images/2026/10/nullg/intro-xl-splitscreen.png)\n\n*Two racers, two views, and independent HUDs in a NullG split-screen session.*\n\nThe longer-term work includes original ships and tracks, sound, and handling for gamepad and simulator use. Motion output, positional audio, and multiple displays remain part of the simulator plan.\n\nThe [NullG project pages](https://projectable.me/projects/nullg/) cover the engine and its tools in more detail. Development footage is available in the [NullG video playlist](https://www.youtube.com/playlist?list=PLaR-_mRLlb6E), with further updates collected in the [project blog](https://projectable.me/projects/nullg/blog/).","html":"\u003cp\u003e\u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG\u003c/a\u003e is an anti-gravity racing engine built in C++20 with Vulkan for rendering. It combines legacy game asset imports with racing physics, weapons, multiplayer, and tools for testing and capturing gameplay.\u003c/p\u003e\n\u003cp\u003eThe longer-term goal is a motion simulator pod, with acceleration, braking, cornering, and impacts feeding into physical movement. Positional audio and multiple displays are part of that plan too. Earlier projects such as the \u003ca href=\"https://projectable.me/my-dirt-rally-tabletop-car-cade-game/\" rel=\"nofollow\"\u003eCar-cade\u003c/a\u003e and \u003ca href=\"https://projectable.me/sim-racing-motion-rig-v2-updates/\" rel=\"nofollow\"\u003emotion rig\u003c/a\u003e explored racing through custom hardware. NullG takes that work into the game engine.\u003c/p\u003e\n\u003cp\u003eDevelopment started in January 2026 with AI-assisted planning and implementation. Over the following months, the work expanded into asset loading, rendering, physics, and automated tests. October added more direct access for agents to inspect the engine, run controlled scenarios, and review the results.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/583bdbe059a6-ship.png\" alt=\"Original NullG ship artwork\"\u003e\n\u003cp\u003e\u003cem\u003eOriginal NullG ship artwork.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eJanuary: plans and the first renderer\u003c/h2\u003e\n\u003cp\u003eThe first git commit, dated January 22, 2026, contains the project outline, research files, architecture notes, and initial application code. The plans already covered asset conversion, input, audio, networking, computer-controlled opponents, and motion output. The application had a window, Vulkan setup, input handling, configuration, and logging.\u003c/p\u003e\n\u003cp\u003eThe architecture put the race simulation at the center. Graphics and sound would use the state produced by the simulation, with motion hardware eventually reading the same data. This kept the simulator goal part of the engine design from the beginning.\u003c/p\u003e\n\u003cp\u003eEarly development used Cursor, with project commands for planning, preparation, and review. The preparation step directed the agent to read reference material and trace the relevant code before making changes. Tasks were broken into smaller pieces, with examples and written decisions available for the next session.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/037075fac8ed-G_jZiq3WQAAg2Cz.jpeg\" alt=\"G_jZiq3WQAAg2Cz.jpeg\"\u003e\n\u003cp\u003eJanuary 23 brought asset conversion tooling and the first renderer pass. Menus, model loading, asset management, and track parsing followed over the next few days. Existing WipEout research and community projects provided references for loading the ships, tracks, textures, and scenery.\u003c/p\u003e\n\u003ch2\u003eFebruary: loading the original assets\u003c/h2\u003e\n\u003cp\u003eBy February 8, the project had a direct PlayStation asset-loading checkpoint. Work during this period covered track rendering, scene objects, sprites, and texture transparency.\u003c/p\u003e\n\u003cp\u003eLoading the geometry, textures, and scenery together made it possible to check the imported data in the renderer. File-format notes and agent rules developed alongside the loaders, giving later tasks a record of how the assets were organized and used.\u003c/p\u003e\n\u003ch2\u003eMarch: handling, race flow, and weapons\u003c/h2\u003e\n\u003cp\u003eMarch focused more on how ships move around a track: hover height, coasting, weight, jumps, boost, recovery, and tunnel-wall collisions. Scene placement, sky textures, pickups, and boost pads also received attention, along with the camera moving through the track behind the menus.\u003c/p\u003e\n\u003cp\u003eReal-time lighting, shadows, and GPU particles added to the presentation. By March 6, the code included race-session flow and a weapon system, handling the transition into a race and back to the menus.\u003c/p\u003e\n\u003cp\u003eThese changes were developed as individual tasks using the relevant code and reference material. Code review and updated implementation notes kept the decisions available as the systems grew.\u003c/p\u003e\n\u003ch2\u003eMay: rendering and automated tests\u003c/h2\u003e\n\u003cp\u003eMay brought a larger pass through the engine internals. The original engine plan was split into focused documents with a shared status index, making individual areas easier to track.\u003c/p\u003e\n\u003cp\u003eRenderer work included dynamic rendering, geometry access, material organization, and a move to Slang for runtime shaders. Lighting work covered animated scene lights, illuminated pickups, shields, and the shadows received by ships from surrounding scenery.\u003c/p\u003e\n\u003cp\u003eAn entity/component scene system and render snapshots were added during the same period. A snapshot holds the scene data needed for a frame, including ship and object positions, giving the renderer a consistent set of data to draw.\u003c/p\u003e\n\u003cp\u003eOn May 28, a headless test suite added checks for the job system, memory, and asset handling. Agents could run these tests after editing the code and inspect the results directly. This became the starting point for the more extensive inspection and gameplay tools added later.\u003c/p\u003e\n\u003ch2\u003eJune: collision, effects, and XL\u003c/h2\u003e\n\u003cp\u003eJune added Jolt collision integration and further GPU effects work. Research notes covered established techniques for effects, shadows, and physics, providing examples for subsequent implementation tasks.\u003c/p\u003e\n\u003cp\u003eWipEout XL became the default asset source in a June 9 checkpoint. Source descriptors drove the menus, allowing the selected game\u0026#39;s data to determine the available choices. Other changes covered waterfall texture scrolling and the starting-grid camera.\u003c/p\u003e\n\u003cp\u003eBy this stage, the codebase covered rendering, asset loading, handling, race flow, weapons, collision, and tests. The next concentrated stretch of commits begins on October 9, with the existing agent guidance brought into Codex, engine-backed inspection tools, and a fixed 60 Hz racing setup. Multiplayer followed on October 10, then the XL reconstruction checkpoint on October 11.\u003c/p\u003e\n\u003ch2\u003eOctober: giving agents access to the engine\u003c/h2\u003e\n\u003cp\u003eOctober also marked a decision to use OpenAI Codex exclusively for NullG\u0026#39;s coding-agent work. The existing planning commands, project rules, and reference notes were carried over from Cursor, keeping the earlier work available as context. Development then focused more heavily on giving Codex access to the tools needed to inspect, test, and review changes.\u003c/p\u003e\n\u003cp\u003eRecently watched YouTube videos about Blender MCP helped shape that approach. They showed how an agent could work directly with a scene and inspect the result. Those concepts were brought into NullG\u0026#39;s model-authoring process and reinforced the value of connecting agents to the engine\u0026#39;s own inspection tools.\u003c/p\u003e\n\u003cp\u003eOctober\u0026#39;s tooling work gave agents more ways to check their changes. \u003ca href=\"https://modelcontextprotocol.io/introduction\" rel=\"nofollow\"\u003eMCP, the Model Context Protocol\u003c/a\u003e, connects AI applications to external tools and data. NullG uses it to expose native inspection windows, track validators, model viewers, and controlled race sessions.\u003c/p\u003e\n\u003cp\u003eAn agent can load a track, choose a camera, inspect geometry, and capture a frame. Model inspection provides access to materials, bounds, and different viewing angles. Test harnesses provide repeatable scenarios with known inputs.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-track-inspection.png\" alt=\"NullG Track Validator showing track geometry, texture inspection, and timeline controls\"\u003e\n\u003cp\u003e\u003cem\u003eThe track inspector, with the track view, texture panel, and playback controls.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe weapon harness opens a paused production race scene. An agent can deploy a weapon and advance the scene in controlled steps, inspecting a projectile, impact, or Quake effect at a specific point. Simulation state and rendered output can then be reviewed together.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/pbUzB3j74p8\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eReactive engine plumes viewed through the MCP harness.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003ePhysics tests compare movement with expected values, track checks inspect road geometry, and screenshots show the rendered result. Replaying the same inputs checks consistency across changes. Together, these tools let agents run a larger part of the validation process themselves, with reports and captures available for human review.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/nullg-ships-weapon-action.png\" alt=\"Ships racing with engine trails and a weapon effect in NullG\"\u003e\n\u003ch3\u003eWorking through a change\u003c/h3\u003e\n\u003cp\u003eA weapon-effect investigation starts with a specific scene and setup. The agent deploys the weapon, steps through the action, checks the state, and follows the code responsible for the effect. After an edit, it builds the affected code and repeats the scenario.\u003c/p\u003e\n\u003cp\u003eSaved reports retain the inputs and settings. That allows the case to be repeated in another session or compared across versions. Visual review and gameplay testing provide additional checks on appearance and handling.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-agent-workflow.png\" alt=\"Diagram showing an agent connected through MCP to native inspection and headless validation tools\"\u003e\n\u003cp\u003e\u003cem\u003eThe agent can run a test, open an inspector, and save the results for review.\u003c/em\u003e\u003c/p\u003e\n\u003ch2\u003eReverse engineering WipEout XL\u003c/h2\u003e\n\u003cp\u003eThe October research also examined WipEout XL / 2097 behavior in more detail, including steering, weapon updates, starting positions, and animated objects.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://github.com/morluto/rea\" rel=\"nofollow\"\u003eREA, Reverse Engineer Anything\u003c/a\u003e, formed part of the research setup. It connects agents with tools for investigating applications and binaries. The PlayStation XL executable decompilation used a PSX-aware Ghidra workflow, with scripts retaining instructions, references, annotations, and findings.\u003c/p\u003e\n\u003cp\u003eThe starting grid provides a useful example. Following the original track numbering and route links established where ships should start. Those findings became checks against the imported track, with the inspector showing the selected position on the road.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-grid-placement.png\" alt=\"An overhead track inspection with a highlighted grid marker\"\u003e\n\u003cp\u003e\u003cem\u003eChecking a grid position against the track geometry.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe research covers both USA and European editions. Each finding retains the edition and executable details so timing values and behavior can be traced to their source.\u003c/p\u003e\n\u003cp\u003eRecovered calculations and state changes are implemented as small C++ reference models, tested, and then connected to NullG. The records distinguish behavior derived from the original executable, adaptations for NullG, and observations from running the engine.\u003c/p\u003e\n\u003cp\u003eAgent access to analysis tools makes this kind of detailed investigation more accessible to individual developers. Existing products can be studied one behavior at a time, with findings carried into tests and new implementations. For NullG, that supports continued refinement of racing behavior alongside modern rendering, controls, and multiplayer. Imported content comes from user-supplied game files, while original NullG assets are developed separately.\u003c/p\u003e\n\u003ch2\u003eA wider wave of software reconstruction\u003c/h2\u003e\n\u003cp\u003eREA\u0026#39;s rapid rise in attention during early October put agent-assisted reverse engineering in front of a much larger audience. \u003ca href=\"https://www.onfirstlaunch.com/blog/reverse-engineer-anything-rea-explained\" rel=\"nofollow\"\u003eCoverage of its growth\u003c/a\u003e described a surge in interest around demonstrations of agents inspecting applications and reconstructing their behavior.\u003c/p\u003e\n\u003cp\u003eAt the same time, “vibe-coded” alternatives to familiar creative and office software were attracting attention. \u003ca href=\"https://github.com/storytold\" rel=\"nofollow\"\u003eArtCraft\u0026#39;s projects\u003c/a\u003e include photo, video, and PDF tools, alongside \u003ca href=\"https://github.com/storytold/wordcraft\" rel=\"nofollow\"\u003eWordCraft\u003c/a\u003e, \u003ca href=\"https://github.com/storytold/gridcraft\" rel=\"nofollow\"\u003eGridCraft\u003c/a\u003e, and \u003ca href=\"https://github.com/storytold/deckcraft\" rel=\"nofollow\"\u003eDeckCraft\u003c/a\u003e, which target Word, Excel, and PowerPoint-style workflows. \u003ca href=\"https://www.techradar.com/computing/office-suites/developer-who-vibe-coded-free-adobe-clones-reveals-apps-to-take-down-microsoft-office-next-and-says-there-is-no-reason-to-save-broken-software\" rel=\"nofollow\"\u003eOctober coverage\u003c/a\u003e connected these AI-assisted projects with the growing interest in alternatives to Adobe and Microsoft applications.\u003c/p\u003e\n\u003cp\u003eThese are separate projects exploring related possibilities: using agents to understand software, implement familiar workflows, and build tools around specific needs. NullG applies that approach to racing. Executable research informs individual behaviors, while tests, inspectors, and recorded scenarios provide ways to check the implementation. The same emphasis on repeatable validation carries through asset production and gameplay development.\u003c/p\u003e\n\u003ch2\u003eBlender and model production\u003c/h2\u003e\n\u003cp\u003eThe Blender MCP concepts from those videos were applied to a specific job: preparing original menu models for the engine. Scene access let the agent inspect parts, make scoped edits, export a candidate, and produce comparison views within the same task.\u003c/p\u003e\n\u003cp\u003eThe benefit was practical. Model changes could be checked from several angles in Blender and then reviewed at their actual size in NullG\u0026#39;s menus. Saved revisions and repeatable exports made the process easier to iterate, while geometry checks and visual review stayed part of the finishing pass.\u003c/p\u003e\n\u003cp\u003eBlender MCP provides agent access to scene inspection, geometry and material editing, and visual review. The menu-model process starts with visual references and generated models where useful, followed by a Blender finishing pass.\u003c/p\u003e\n\u003cp\u003eThat pass covers shape, individual parts, surface cleanup, materials, and polygon count. Models are reviewed from several angles, exported, and checked in NullG\u0026#39;s model viewer and menus at their intended display size.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-blender-graphics-model.png\" alt=\"Four Blender review renders of the original monitor-shaped Graphics menu model\"\u003e\n\u003cp\u003e\u003cem\u003eFront, side, rear, and angled review renders of the Graphics menu model.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe authoring files, exports, and comparison images remain available through each revision. Connecting these tools lets an agent work through research, code, tests, and visual checks within the same development session.\u003c/p\u003e\n\u003ch2\u003eRace capture and social media\u003c/h2\u003e\n\u003cp\u003eNullG includes race recording, replay playback, cinematic cameras, and video export. Recorded races can be rendered with ship-focused shots, action cuts, titles, and audio. Landscape, portrait, and square formats support both longer videos and shorter social clips.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/tREm0SM8zmg\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eThe XL cinematic race intro.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eScreenshot tools capture menus, starting grids, tracks, and weapons in action. Agents can help select and organize those captures, with settings and source details retained alongside the images. The same recorded race can support a gameplay review and provide footage for a project update.\u003c/p\u003e\n\u003ch2\u003eReinforcement learning for the pilots\u003c/h2\u003e\n\u003cp\u003eNullG also includes reinforcement-learning experiments for computer-controlled pilots. A trained model selects ship controls during a race. Development agents help build the training tools, run checks, and prepare the results for review.\u003c/p\u003e\n\u003cp\u003eThe pilot receives information about speed and direction, the road ahead, nearby ships, and pickups. It selects throttle, steering, air brakes, pitch, and weapon commands, which pass through the racing simulation as player-style inputs.\u003c/p\u003e\n\u003cp\u003eTraining uses Python, Gymnasium, Stable-Baselines3, and PyTorch, with PPO as the learning algorithm. Native workers run NullG\u0026#39;s race simulation, allowing several driving attempts to run in parallel while the training process updates the model.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-rl-city-race.png\" alt=\"\"\u003e\n\u003cp\u003e\u003cem\u003eA frame from an RL evaluation replay on Gare d’Europa, rendered through NullG\u0026#39;s replay tools.\u003c/em\u003e\u003c/p\u003e\n\u003ch3\u003eTraining and evaluation\u003c/h3\u003e\n\u003cp\u003eA short run checks the training setup and controls. Longer runs save checkpoints for evaluation on a chosen set of tracks and race configurations. Saved settings and seeds allow those checks to be repeated.\u003c/p\u003e\n\u003cp\u003eEvaluation measures completion, time, damage, and steering smoothness. Traffic and combat scenarios also examine race position, pickup use, and weapon timing. Training checkpoints and policies approved for gameplay have separate evaluation steps.\u003c/p\u003e\n\u003cp\u003eReplays provide another view of the results. Chase cameras show ship movement through a corner, while overhead views show racer positions around the circuit. These views can be compared with the recorded controls and race metrics.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-rl-overhead.png\" alt=\"\"\u003e\n\u003cp\u003e\u003cem\u003eAn overhead replay view from the RL development work, with colored racer markers.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe exported pilot is a compact model evaluated by native C++ code. Python handles training, and the game runs the exported policy. Retained checkpoints allow different training runs to be compared using the same evaluation setup.\u003c/p\u003e\n\u003cp\u003eAgents can run the checks, read reports, and prepare replay captures. This supports review of individual driving behaviors and produces footage for documenting the experiments.\u003c/p\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/fkUTwW2p_8k\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003cp\u003e\u003cem\u003eAn AI-driving training clip from the NullG playlist.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eMore detail is available on the \u003ca href=\"https://projectable.me/projects/nullg/mcp-tooling/\" rel=\"nofollow\"\u003eMCP tooling page\u003c/a\u003e and \u003ca href=\"https://projectable.me/projects/nullg/reinforcement-learning/\" rel=\"nofollow\"\u003ereinforcement-learning page\u003c/a\u003e.\u003c/p\u003e\n\u003ch2\u003ePublishing to projectable.me\u003c/h2\u003e\n\u003cp\u003eThe same agent-assisted workflow extends to publishing on projectable.me. Agents can prepare illustrated posts, update project pages, and organize development updates under the NullG project.\u003c/p\u003e\n\u003cp\u003eThe process starts with the git history, project documentation, and relevant captures. An agent prepares the article, adds screenshots and video embeds, and saves a draft on projectable.me. The post can then be reviewed and edited on the site before publication.\u003c/p\u003e\n\u003cp\u003eThis post follows that process, using repository history for the timeline, project documentation for the workflow details, and existing engine captures for the screenshots. Keeping these updates attached to the NullG project makes the development history easy to follow alongside the project pages.\u003c/p\u003e\n\u003ch2\u003eCurrent state and next steps\u003c/h2\u003e\n\u003cp\u003eNullG now includes racing, weapons, computer-controlled opponents, local split-screen, LAN and private online rooms, and replay tools. Inspection and validation tools support development across those systems.\u003c/p\u003e\n\u003cimg src=\"/content/images/2026/10/nullg/intro-xl-splitscreen.png\" alt=\"Two-player vertical split-screen racing with WipEout XL content in NullG\"\u003e\n\u003cp\u003e\u003cem\u003eTwo racers, two views, and independent HUDs in a NullG split-screen session.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe longer-term work includes original ships and tracks, sound, and handling for gamepad and simulator use. Motion output, positional audio, and multiple displays remain part of the simulator plan.\u003c/p\u003e\n\u003cp\u003eThe \u003ca href=\"https://projectable.me/projects/nullg/\" rel=\"nofollow\"\u003eNullG project pages\u003c/a\u003e cover the engine and its tools in more detail. Development footage is available in the \u003ca href=\"https://www.youtube.com/playlist?list=PLaR-_mRLlb6E\" rel=\"nofollow\"\u003eNullG video playlist\u003c/a\u003e, with further updates collected in the \u003ca href=\"https://projectable.me/projects/nullg/blog/\" rel=\"nofollow\"\u003eproject blog\u003c/a\u003e.\u003c/p\u003e\n","excerpt":"NullG’s development from January’s engine plans through asset loading, racing systems, XL research, agent validation, and reinforcement learning.","metaTitle":"Introducing NullG: racing and AI-assisted development","metaDescription":"The development of NullG: a C++20 and Vulkan racing engine with WipEout XL research, MCP tools, Blender workflows, reinforcement learning, and replay capture.","status":"published","featured":true,"publishedAt":"2026-10-19T02:27:51Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"3cd28585400f94ee7115b744db06b6a1","projectId":"ae4bdb66da6dc1729a263a475e1cdbe3","tags":[{"id":"1c9e8a1339a8f8b797767e91a9673016","name":"Game","slug":"game"},{"id":"ecaaace3f367d1a9865ae0933d53d715","name":"Sim Racing","slug":"sim-racing"}],"createdAt":"2026-10-19T02:05:13Z","updatedAt":"2026-10-19T20:15:35Z"},{"id":"add9e77292c614f59260072375d13db5","siteId":"projectable","slug":"ubuntu-24-04-nvidia-drivers-ollama","title":"Ubuntu 24.04 + Nvidia Drivers + Ollama","markdown":"![](/content/images/2025/03/IMG_8241.JPG)\n\nI am building an updated server to play with LLMs for local home automation control using Home Assistant (https://www.home-assistant.io). \n\nI managed to pick up a Dell R740 for a great price ($400) and with a few additional eBay purchases have a nice server capable of housing 3 GPUs. I have an existing Nvidia Tesla P100 from my other server that I am moving to this new server, and will eventually look at finding a Tesla V100 32G PCIe as they seem to have a slightly more accessible price compared to the Tesla A100 PCIe. \n\n![](/content/images/2025/03/IMG_8212.JPG)\n![](/content/images/2025/03/IMG_8213.JPG)\n![](/content/images/2025/03/IMG_8216.jpg)\n\nI will continue to run Home Assistant on my [Home Assistant Yellow](https://www.home-assistant.io/yellow/) (PoE, Raspberry Pi CM5 8GB, NVMe SSD), but will land Ollama, and other tools on this new server.\n\n![](/content/images/2025/03/IMG_8243.JPG)\n\nI was banging my head against this for the 2nd time I thought it would be helpful to document the setup process for getting Ollama running on Ubuntu 24.04 along with the steps to reliably install the Nvidia drivers \u0026 CUDA packages. \n\n\n### Nvidia Driver install\n\nBefore you begin, ensure that your Nvidia card is showing up on the PCI bus. Run `lspci | grep -i nvidia` and you should get output similar to this:\n```\naf:00.0 3D controller: NVIDIA Corporation GP100GL [Tesla P100 PCIe 16GB] (rev a1)\n```\n\nIf you have already attempted to install nvidia drivers and have failed cleanup your existing Nvidia nuggets as follows:\n```sh\nsudo apt-get remove --purge 'libnvidia-.*'\nsudo apt-get remove --purge '^nvidia-.*'\nsudo apt-get remove --purge '^libnvidia-.*'\nsudo apt-get remove --purge '^cuda-.*'\nsudo apt clean\nsudo apt autoremove\n```\n\nNow install the updated Nvidia drivers:\n```sh\nsudo add-apt-repository ppa:graphics-drivers/ppa --yes\nsudo apt-get update\n\nupdate-pciids\n\nsudo apt-get install nvidia-driver-570 -y\nsudo apt-get reinstall linux-headers-$(uname -r)\n\nsudo update-initramfs -u\n```\n\nValidate that the DKMS modules are actually there, and then reboot if they are:\n```sh\nsudo dkms status\nsudo reboot\n```\n\nAfter a reboot you should be able to run `nvidia-smi` and get output similar to this:\n```\nroot@r740:~# nvidia-smi \nWed Mar 5 14:34:02 2025 \n+-----------------------------------------------------------------------------------------+\n| NVIDIA-SMI 570.86.16 Driver Version: 570.86.16 CUDA Version: 12.8 |\n|-----------------------------------------+------------------------+----------------------+\n| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |\n| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |\n| | | MIG M. |\n|=========================================+========================+======================|\n| 0 Tesla P100-PCIE-16GB Off | 00000000:AF:00.0 Off | 0 |\n| N/A 30C P0 26W / 250W | 0MiB / 16384MiB | 0% Default |\n| | | N/A |\n+-----------------------------------------+------------------------+----------------------+\n \n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| No running processes found |\n+-----------------------------------------------------------------------------------------+\n```\n\n\n### Nvidia CUDA install\n\n```sh\nwget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb\nsudo dpkg -i cuda-keyring_1.1-1_all.deb\n\nsudo apt-get update\n\nsudo apt-get install cuda-toolkit -y\nsudo apt-get install nvidia-gds -y\n```\n\nI don't think the `nvidia-gds` package is required for my P100, but the official CUDA install guide has it so I'm including it here. Check that the `nvidia-fs` module is now available in DKMS and then reboot.\n\n```sh\nsudo dkms status\nsudo reboot\n```\n\nAfter rebooting you will likely see that there are updated packages for the Nvidia drivers, you can upgrade them if you like. I found that installing the drivers from the `ppa:graphics-drivers/ppa` repository first had more consistent results than installing from the nvidia repos first. Your mileage may vary... \n\n\n### Install Docker\nWe will use docker to host the ollama web ui, but we will run ollama locally on the system.\n```sh\nsudo apt-get install curl apt-transport-https ca-certificates software-properties-common -y\n\ncurl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg\necho \"deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable\" | sudo tee /etc/apt/sources.list.d/docker.list \u003e /dev/null\n\nsudo apt-get update\nsudo apt-get install docker-ce -y\n```\nNow add your local user to the docker group `sudo usermod -aG docker $USER`.\n\n\n### Install Nvidia Container Toolkit\n```sh\ncurl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \\\n \u0026\u0026 curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \\\n sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \\\n sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list\n\nsudo apt-get update\nsudo apt-get install nvidia-container-toolkit -y\n\nsudo systemctl restart docker\n```\n\n### Install Ollama\n```sh\ncurl -fsSL https://ollama.com/install.sh | sh\n```\n\nNow edit the systemd service file to allow the Ollama to listen on all network interfaces on the system instead of just `localhost` by adding `OLLAMA_HOST=0.0.0.0:11434`\n\n**Edit**: `/etc/systemd/system/ollama.service`\n```\n[Unit]\nDescription=Ollama Service\nAfter=network-online.target\n\n[Service]\nExecStart=/usr/local/bin/ollama serve\nUser=ollama\nGroup=ollama\nRestart=always\nRestartSec=3\nEnvironment=\"PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin\"\nEnvironment=\"OLLAMA_HOST=0.0.0.0:11434\"\n\n[Install]\nWantedBy=default.target\n```\n\nThen run the following to apply the changes:\n```sh\nsudo systemctl daemon-reload\nsudo systemctl restart ollama \n```\n\n\n### Install Open WebUI\n```sh\ndocker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main\n```\n\nYou should now be able to navigate to Open WebUI using your server IP and port 3000. Create your admin account and you are ready to start installing models from https://ollama.com/library.\n\n![](/content/images/2025/03/ollama1-1.png)\n![](/content/images/2025/03/ollama2.png)","html":"\u003cimg src=\"/content/images/2025/03/IMG_8241.JPG\" alt=\"\"\u003e\n\u003cp\u003eI am building an updated server to play with LLMs for local home automation control using Home Assistant (\u003ca href=\"https://www.home-assistant.io\" rel=\"nofollow\"\u003ehttps://www.home-assistant.io\u003c/a\u003e).\u003c/p\u003e\n\u003cp\u003eI managed to pick up a Dell R740 for a great price ($400) and with a few additional eBay purchases have a nice server capable of housing 3 GPUs. I have an existing Nvidia Tesla P100 from my other server that I am moving to this new server, and will eventually look at finding a Tesla V100 32G PCIe as they seem to have a slightly more accessible price compared to the Tesla A100 PCIe.\u003c/p\u003e\n\u003cimg src=\"/content/images/2025/03/IMG_8212.JPG\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2025/03/IMG_8213.JPG\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2025/03/IMG_8216.jpg\" alt=\"\"\u003e\n\u003cp\u003eI will continue to run Home Assistant on my \u003ca href=\"https://www.home-assistant.io/yellow/\" rel=\"nofollow\"\u003eHome Assistant Yellow\u003c/a\u003e (PoE, Raspberry Pi CM5 8GB, NVMe SSD), but will land Ollama, and other tools on this new server.\u003c/p\u003e\n\u003cimg src=\"/content/images/2025/03/IMG_8243.JPG\" alt=\"\"\u003e\n\u003cp\u003eI was banging my head against this for the 2nd time I thought it would be helpful to document the setup process for getting Ollama running on Ubuntu 24.04 along with the steps to reliably install the Nvidia drivers \u0026amp; CUDA packages.\u003c/p\u003e\n\u003ch3\u003eNvidia Driver install\u003c/h3\u003e\n\u003cp\u003eBefore you begin, ensure that your Nvidia card is showing up on the PCI bus. Run \u003ccode\u003elspci | grep -i nvidia\u003c/code\u003e and you should get output similar to this:\u003c/p\u003e\n\u003cpre\u003e\u003ccode\u003eaf:00.0 3D controller: NVIDIA Corporation GP100GL [Tesla P100 PCIe 16GB] (rev a1)\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eIf you have already attempted to install nvidia drivers and have failed cleanup your existing Nvidia nuggets as follows:\u003c/p\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003esudo apt-get remove --purge \u0026#39;libnvidia-.*\u0026#39;\nsudo apt-get remove --purge \u0026#39;^nvidia-.*\u0026#39;\nsudo apt-get remove --purge \u0026#39;^libnvidia-.*\u0026#39;\nsudo apt-get remove --purge \u0026#39;^cuda-.*\u0026#39;\nsudo apt clean\nsudo apt autoremove\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eNow install the updated Nvidia drivers:\u003c/p\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003esudo add-apt-repository ppa:graphics-drivers/ppa --yes\nsudo apt-get update\n\nupdate-pciids\n\nsudo apt-get install nvidia-driver-570 -y\nsudo apt-get reinstall linux-headers-$(uname -r)\n\nsudo update-initramfs -u\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eValidate that the DKMS modules are actually there, and then reboot if they are:\u003c/p\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003esudo dkms status\nsudo reboot\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eAfter a reboot you should be able to run \u003ccode\u003envidia-smi\u003c/code\u003e and get output similar to this:\u003c/p\u003e\n\u003cpre\u003e\u003ccode\u003eroot@r740:~# nvidia-smi \nWed Mar 5 14:34:02 2025 \n+-----------------------------------------------------------------------------------------+\n| NVIDIA-SMI 570.86.16 Driver Version: 570.86.16 CUDA Version: 12.8 |\n|-----------------------------------------+------------------------+----------------------+\n| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |\n| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |\n| | | MIG M. |\n|=========================================+========================+======================|\n| 0 Tesla P100-PCIE-16GB Off | 00000000:AF:00.0 Off | 0 |\n| N/A 30C P0 26W / 250W | 0MiB / 16384MiB | 0% Default |\n| | | N/A |\n+-----------------------------------------+------------------------+----------------------+\n \n+-----------------------------------------------------------------------------------------+\n| Processes: |\n| GPU GI CI PID Type Process name GPU Memory |\n| ID ID Usage |\n|=========================================================================================|\n| No running processes found |\n+-----------------------------------------------------------------------------------------+\n\u003c/code\u003e\u003c/pre\u003e\n\u003ch3\u003eNvidia CUDA install\u003c/h3\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003ewget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb\nsudo dpkg -i cuda-keyring_1.1-1_all.deb\n\nsudo apt-get update\n\nsudo apt-get install cuda-toolkit -y\nsudo apt-get install nvidia-gds -y\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eI don\u0026#39;t think the \u003ccode\u003envidia-gds\u003c/code\u003e package is required for my P100, but the official CUDA install guide has it so I\u0026#39;m including it here. Check that the \u003ccode\u003envidia-fs\u003c/code\u003e module is now available in DKMS and then reboot.\u003c/p\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003esudo dkms status\nsudo reboot\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eAfter rebooting you will likely see that there are updated packages for the Nvidia drivers, you can upgrade them if you like. I found that installing the drivers from the \u003ccode\u003eppa:graphics-drivers/ppa\u003c/code\u003e repository first had more consistent results than installing from the nvidia repos first. Your mileage may vary...\u003c/p\u003e\n\u003ch3\u003eInstall Docker\u003c/h3\u003e\n\u003cp\u003eWe will use docker to host the ollama web ui, but we will run ollama locally on the system.\u003c/p\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003esudo apt-get install curl apt-transport-https ca-certificates software-properties-common -y\n\ncurl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /usr/share/keyrings/docker-archive-keyring.gpg\necho \u0026#34;deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/docker-archive-keyring.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable\u0026#34; | sudo tee /etc/apt/sources.list.d/docker.list \u0026gt; /dev/null\n\nsudo apt-get update\nsudo apt-get install docker-ce -y\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eNow add your local user to the docker group \u003ccode\u003esudo usermod -aG docker $USER\u003c/code\u003e.\u003c/p\u003e\n\u003ch3\u003eInstall Nvidia Container Toolkit\u003c/h3\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003ecurl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \\\n \u0026amp;\u0026amp; curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \\\n sed \u0026#39;s#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g\u0026#39; | \\\n sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list\n\nsudo apt-get update\nsudo apt-get install nvidia-container-toolkit -y\n\nsudo systemctl restart docker\n\u003c/code\u003e\u003c/pre\u003e\n\u003ch3\u003eInstall Ollama\u003c/h3\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003ecurl -fsSL https://ollama.com/install.sh | sh\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eNow edit the systemd service file to allow the Ollama to listen on all network interfaces on the system instead of just \u003ccode\u003elocalhost\u003c/code\u003e by adding \u003ccode\u003eOLLAMA_HOST=0.0.0.0:11434\u003c/code\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEdit\u003c/strong\u003e: \u003ccode\u003e/etc/systemd/system/ollama.service\u003c/code\u003e\u003c/p\u003e\n\u003cpre\u003e\u003ccode\u003e[Unit]\nDescription=Ollama Service\nAfter=network-online.target\n\n[Service]\nExecStart=/usr/local/bin/ollama serve\nUser=ollama\nGroup=ollama\nRestart=always\nRestartSec=3\nEnvironment=\u0026#34;PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin\u0026#34;\nEnvironment=\u0026#34;OLLAMA_HOST=0.0.0.0:11434\u0026#34;\n\n[Install]\nWantedBy=default.target\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eThen run the following to apply the changes:\u003c/p\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003esudo systemctl daemon-reload\nsudo systemctl restart ollama \n\u003c/code\u003e\u003c/pre\u003e\n\u003ch3\u003eInstall Open WebUI\u003c/h3\u003e\n\u003cpre\u003e\u003ccode class=\"language-sh\"\u003edocker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eYou should now be able to navigate to Open WebUI using your server IP and port 3000. Create your admin account and you are ready to start installing models from \u003ca href=\"https://ollama.com/library\" rel=\"nofollow\"\u003ehttps://ollama.com/library\u003c/a\u003e.\u003c/p\u003e\n\u003cimg src=\"/content/images/2025/03/ollama1-1.png\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2025/03/ollama2.png\" alt=\"\"\u003e","excerpt":"","metaTitle":"","metaDescription":"","status":"published","featured":false,"publishedAt":"2026-10-18T18:03:14Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"","projectId":"","tags":[],"createdAt":"2026-10-18T18:03:08Z","updatedAt":"2026-10-18T18:03:14Z"},{"id":"729c9a4edc92eaa4171ac274dda70ca2","siteId":"projectable","slug":"comcast-business-ipv6-on-ubiquiti-unifi","title":"Comcast Business IPv6 on Ubiquiti Unifi","markdown":"If you are reading this it is likely you were like me, on the edge of giving up on IPv6 on your network. After combining a few small pieces from posts outlining **Comcast Business IPv6** using Ubiquiti EdgeRouters along with the [**USG Advanced Configuration**](https://help.ubnt.com/hc/en-us/articles/215458888-UniFi-USG-Advanced-Configuration) guide I was able to get something that not only works correctly, but persists correctly also.\n\nThe trick was getting the right settings defined in the web interface and the one setting not available in the web interface. It's pretty straight forward, so for sake of simplicity I will just tell you what things need to be set. *This configuration has 1 of my internal networks having IPv6, I did not want to spend any more time trying to make this work across multiple internal networks, but it probably isn't hard.*\n\n#### WAN Interface\n![](/content/images/2019/11/Screen-Shot-2019-11-05-at-7-49-35-AM.png)\n\nSet your WAN interface **IPv6 Connection Type** to `Using DHCPv6`. You will need to ensure that your cable modem has DHCPv6 enabled, it usually is by default. The **Prefix Delegation Size** should be set to `59` as 56 didn't seem to work.\n\n\n### LAN Interface\n![](/content/images/2019/11/Screen-Shot-2019-11-05-at-7-52-39-AM.png)\n\n**IPv6 Interface Type** gets set to `Prefix Delegation`, you select the WAN interface you used earlier, and you set the **IPv6 Prefix ID** to `1`. \n\nI suggest setting **DHCPv6/RDNSS DNS Control** to `Auto`, mine is set to `Manual` due to having a [Pi-hole](https://pi-hole.net) on my network that I use for DNS recursion/caching.\n\n\n### The Backend Stuff\nYou will need to follow the **USG Advanced Configuration** guide found here: https://help.ubnt.com/hc/en-us/articles/215458888-UniFi-USG-Advanced-Configuration.\n\nFollowing that guide you create your own `config.gateway.json` file and can paste the following.\n```\n{\n \"interfaces\": {\n \"ethernet\": {\n \"eth2\": {\n \"dhcpv6-pd\": {\n \"rapid-commit\": \"disable\"\n }\n }\n }\n }\n}\n```\nIf you are using **WAN2** instead of **WAN** change `eth2` to `eth3`. \n\n### Finishing Up\nIf you followed the guide and **Force provision**ed your **USG** you should be good to go. \nMy linux hosts picked up the IPv6 IPs automatically, and on my Apple products I just disabled then re-enabled Wifi.","html":"\u003cp\u003eIf you are reading this it is likely you were like me, on the edge of giving up on IPv6 on your network. After combining a few small pieces from posts outlining \u003cstrong\u003eComcast Business IPv6\u003c/strong\u003e using Ubiquiti EdgeRouters along with the \u003ca href=\"https://help.ubnt.com/hc/en-us/articles/215458888-UniFi-USG-Advanced-Configuration\" rel=\"nofollow\"\u003e\u003cstrong\u003eUSG Advanced Configuration\u003c/strong\u003e\u003c/a\u003e guide I was able to get something that not only works correctly, but persists correctly also.\u003c/p\u003e\n\u003cp\u003eThe trick was getting the right settings defined in the web interface and the one setting not available in the web interface. It\u0026#39;s pretty straight forward, so for sake of simplicity I will just tell you what things need to be set. \u003cem\u003eThis configuration has 1 of my internal networks having IPv6, I did not want to spend any more time trying to make this work across multiple internal networks, but it probably isn\u0026#39;t hard.\u003c/em\u003e\u003c/p\u003e\n\u003ch4\u003eWAN Interface\u003c/h4\u003e\n\u003cimg src=\"/content/images/2019/11/Screen-Shot-2019-11-05-at-7-49-35-AM.png\" alt=\"\"\u003e\n\u003cp\u003eSet your WAN interface \u003cstrong\u003eIPv6 Connection Type\u003c/strong\u003e to \u003ccode\u003eUsing DHCPv6\u003c/code\u003e. You will need to ensure that your cable modem has DHCPv6 enabled, it usually is by default. The \u003cstrong\u003ePrefix Delegation Size\u003c/strong\u003e should be set to \u003ccode\u003e59\u003c/code\u003e as 56 didn\u0026#39;t seem to work.\u003c/p\u003e\n\u003ch3\u003eLAN Interface\u003c/h3\u003e\n\u003cimg src=\"/content/images/2019/11/Screen-Shot-2019-11-05-at-7-52-39-AM.png\" alt=\"\"\u003e\n\u003cp\u003e\u003cstrong\u003eIPv6 Interface Type\u003c/strong\u003e gets set to \u003ccode\u003ePrefix Delegation\u003c/code\u003e, you select the WAN interface you used earlier, and you set the \u003cstrong\u003eIPv6 Prefix ID\u003c/strong\u003e to \u003ccode\u003e1\u003c/code\u003e.\u003c/p\u003e\n\u003cp\u003eI suggest setting \u003cstrong\u003eDHCPv6/RDNSS DNS Control\u003c/strong\u003e to \u003ccode\u003eAuto\u003c/code\u003e, mine is set to \u003ccode\u003eManual\u003c/code\u003e due to having a \u003ca href=\"https://pi-hole.net\" rel=\"nofollow\"\u003ePi-hole\u003c/a\u003e on my network that I use for DNS recursion/caching.\u003c/p\u003e\n\u003ch3\u003eThe Backend Stuff\u003c/h3\u003e\n\u003cp\u003eYou will need to follow the \u003cstrong\u003eUSG Advanced Configuration\u003c/strong\u003e guide found here: \u003ca href=\"https://help.ubnt.com/hc/en-us/articles/215458888-UniFi-USG-Advanced-Configuration\" rel=\"nofollow\"\u003ehttps://help.ubnt.com/hc/en-us/articles/215458888-UniFi-USG-Advanced-Configuration\u003c/a\u003e.\u003c/p\u003e\n\u003cp\u003eFollowing that guide you create your own \u003ccode\u003econfig.gateway.json\u003c/code\u003e file and can paste the following.\u003c/p\u003e\n\u003cpre\u003e\u003ccode\u003e{\n \u0026#34;interfaces\u0026#34;: {\n \u0026#34;ethernet\u0026#34;: {\n \u0026#34;eth2\u0026#34;: {\n \u0026#34;dhcpv6-pd\u0026#34;: {\n \u0026#34;rapid-commit\u0026#34;: \u0026#34;disable\u0026#34;\n }\n }\n }\n }\n}\n\u003c/code\u003e\u003c/pre\u003e\n\u003cp\u003eIf you are using \u003cstrong\u003eWAN2\u003c/strong\u003e instead of \u003cstrong\u003eWAN\u003c/strong\u003e change \u003ccode\u003eeth2\u003c/code\u003e to \u003ccode\u003eeth3\u003c/code\u003e.\u003c/p\u003e\n\u003ch3\u003eFinishing Up\u003c/h3\u003e\n\u003cp\u003eIf you followed the guide and \u003cstrong\u003eForce provision\u003c/strong\u003eed your \u003cstrong\u003eUSG\u003c/strong\u003e you should be good to go.\nMy linux hosts picked up the IPv6 IPs automatically, and on my Apple products I just disabled then re-enabled Wifi.\u003c/p\u003e\n","excerpt":"","metaTitle":"","metaDescription":"","status":"published","featured":false,"publishedAt":"2019-11-05T13:07:15Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"a89f927e3c656e9526008e3aee6a678d","projectId":"","tags":[{"id":"d35708b28e494df332afc857416da120","name":"Unifi","slug":"unifi"},{"id":"75282f5450b7d8d38ca5570b768f4816","name":"Ubiquiti","slug":"ubiquiti"},{"id":"ef5b5edfc8d3f826a59d61a4be854fd8","name":"Networking","slug":"networking"}],"createdAt":"2026-10-18T13:16:27Z","updatedAt":"2026-10-18T13:16:27Z"},{"id":"901f2fab2646e83671fe8f6e5500c117","siteId":"projectable","slug":"a-simple-etcher-6-slot-sd-card-jig","title":"A simple Etcher 6 slot SD card jig","markdown":"I've been working with Kubernetes a lot lately, I am using Kubernetes at [Filament](https://filament.com) as a way to manage sandbox blockchain environments, so naturally I wanted to try and replicate this environment using Raspberry Pis. This is actually easier said than done, even though the 3B/3B+ has 1GB of RAM, once you start bringing in persistent storage layers it becomes apparent that it just isn't enough memory to run anything but the simplest of kubernetes containers.\n\n![](/content/images/2018/11/IMG_6776.jpg)\n\nThroughout the testing and validation of Kubernetes on my 6 node 3B cluster I was constantly re-imaging the SD cards. I tested Raspbian and Ubuntu 18 64bit, with various full resets along the way. The process of imaging 6 cards with only 1 SD reader is a bit monotonous, so I set out to find out what options there were for multiple card writing systems. I was very quickly overwhelmed by the $1500+ SD card duplication systems out there and I realized I would have to cobble something together myself.\n\n####Doing the jig\n\nA quick search on Amazon and I found a [USB 3.0 dual SD/Micro SD card reader](http://a.co/d/fkdFMf3), I figured I could get 3 of these and plug them in to a hub to give me the 6 card slots I needed. I was then pleasantly surprised to find the manufacturer of the card readers also had a [4 port USB 3.0 hub](http://a.co/d/bhqes7W) that was the same size! Once I got all the hardware in hand I could see that the hub was almost exactly the same base dimensions as the 3 dual card readers stacked next to each other!\n\n![](/content/images/2018/11/IMG_6730.jpg)\n![](/content/images/2018/11/IMG_6731.jpg)\n\nI designed a simple 3D printable enclosure in Fusion360 and then printed it using Proto-Pasta Metallic Blue HTPLA. It is held together using M3 hardware.\n\n![](/content/images/2018/11/Screen-Shot-2018-11-19-at-12-50-13-PM.png)\n\nThe finished enclosure keeps everything together, and doesn't look half bad!\n\n![](/content/images/2018/11/IMG_6733.jpg)\n![](/content/images/2018/11/IMG_6735.jpg)\n\nThis setup gives you 6 usable SD/Micro SD card slots with 1 USB 3.0 cable, and a spare USB 3.0 port to daisy-chain more of these jigs if needed! [Etcher](https://www.balena.io/etcher/) can write to all of the SD cards at the same time, cutting the burn time significantly!\n\n![](/content/images/2018/11/Screen-Shot-2018-11-13-at-7-58-50-AM.png)\n\nA big thanks to [Balena](https://www.balena.io) (previously Resin.io) for releasing [Etcher](https://www.balena.io/etcher/) to the world, its a fantastic tool that scales well!\n\n####Go make one yourself!\nAll of the STLs and the BOM can be found on Thingiverse: https://www.thingiverse.com/thing:3217069\n\n\n","html":"\u003cp\u003eI\u0026#39;ve been working with Kubernetes a lot lately, I am using Kubernetes at \u003ca href=\"https://filament.com\" rel=\"nofollow\"\u003eFilament\u003c/a\u003e as a way to manage sandbox blockchain environments, so naturally I wanted to try and replicate this environment using Raspberry Pis. This is actually easier said than done, even though the 3B/3B+ has 1GB of RAM, once you start bringing in persistent storage layers it becomes apparent that it just isn\u0026#39;t enough memory to run anything but the simplest of kubernetes containers.\u003c/p\u003e\n\u003cimg src=\"/content/images/2018/11/IMG_6776.jpg\" alt=\"\"\u003e\n\u003cp\u003eThroughout the testing and validation of Kubernetes on my 6 node 3B cluster I was constantly re-imaging the SD cards. I tested Raspbian and Ubuntu 18 64bit, with various full resets along the way. The process of imaging 6 cards with only 1 SD reader is a bit monotonous, so I set out to find out what options there were for multiple card writing systems. I was very quickly overwhelmed by the $1500+ SD card duplication systems out there and I realized I would have to cobble something together myself.\u003c/p\u003e\n\u003ch4\u003eDoing the jig\u003c/h4\u003e\n\u003cp\u003eA quick search on Amazon and I found a \u003ca href=\"http://a.co/d/fkdFMf3\" rel=\"nofollow\"\u003eUSB 3.0 dual SD/Micro SD card reader\u003c/a\u003e, I figured I could get 3 of these and plug them in to a hub to give me the 6 card slots I needed. I was then pleasantly surprised to find the manufacturer of the card readers also had a \u003ca href=\"http://a.co/d/bhqes7W\" rel=\"nofollow\"\u003e4 port USB 3.0 hub\u003c/a\u003e that was the same size! Once I got all the hardware in hand I could see that the hub was almost exactly the same base dimensions as the 3 dual card readers stacked next to each other!\u003c/p\u003e\n\u003cimg src=\"/content/images/2018/11/IMG_6730.jpg\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2018/11/IMG_6731.jpg\" alt=\"\"\u003e\n\u003cp\u003eI designed a simple 3D printable enclosure in Fusion360 and then printed it using Proto-Pasta Metallic Blue HTPLA. It is held together using M3 hardware.\u003c/p\u003e\n\u003cimg src=\"/content/images/2018/11/Screen-Shot-2018-11-19-at-12-50-13-PM.png\" alt=\"\"\u003e\n\u003cp\u003eThe finished enclosure keeps everything together, and doesn\u0026#39;t look half bad!\u003c/p\u003e\n\u003cimg src=\"/content/images/2018/11/IMG_6733.jpg\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2018/11/IMG_6735.jpg\" alt=\"\"\u003e\n\u003cp\u003eThis setup gives you 6 usable SD/Micro SD card slots with 1 USB 3.0 cable, and a spare USB 3.0 port to daisy-chain more of these jigs if needed! \u003ca href=\"https://www.balena.io/etcher/\" rel=\"nofollow\"\u003eEtcher\u003c/a\u003e can write to all of the SD cards at the same time, cutting the burn time significantly!\u003c/p\u003e\n\u003cimg src=\"/content/images/2018/11/Screen-Shot-2018-11-13-at-7-58-50-AM.png\" alt=\"\"\u003e\n\u003cp\u003eA big thanks to \u003ca href=\"https://www.balena.io\" rel=\"nofollow\"\u003eBalena\u003c/a\u003e (previously Resin.io) for releasing \u003ca href=\"https://www.balena.io/etcher/\" rel=\"nofollow\"\u003eEtcher\u003c/a\u003e to the world, its a fantastic tool that scales well!\u003c/p\u003e\n\u003ch4\u003eGo make one yourself!\u003c/h4\u003e\n\u003cp\u003eAll of the STLs and the BOM can be found on Thingiverse: \u003ca href=\"https://www.thingiverse.com/thing:3217069\" rel=\"nofollow\"\u003ehttps://www.thingiverse.com/thing:3217069\u003c/a\u003e\u003c/p\u003e\n","excerpt":"","metaTitle":"","metaDescription":"","status":"published","featured":false,"publishedAt":"2018-11-19T17:53:22Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"453a5a37a265940047559741a496164e","projectId":"","tags":[{"id":"b529314e428273696b02c474dcf56132","name":"Raspberry Pi","slug":"raspberry-pi"},{"id":"1e4c1c80a23c5c7fa3f6f43985fffb50","name":"Kubernetes","slug":"kubernetes"},{"id":"41dfeb58d853ede0e7f6fb6f9e24536c","name":"Etcher","slug":"etcher"},{"id":"e58b4788e99c9f53befd1e47b8073250","name":"3D Printing","slug":"3d-printing"}],"createdAt":"2026-10-18T13:16:27Z","updatedAt":"2026-10-18T13:16:27Z"},{"id":"7a23b4fc9f63f7327ef52913289d4327","siteId":"projectable","slug":"sim-racing-motion-rig-v2-updates","title":"Sim racing motion rig v2 updates","markdown":"With the rekindled interest in Dirt Rally that my [Car-cade](http://projectable.me/my-dirt-rally-tabletop-car-cade-game/) has given me I had a bit of a yearning to get Dirt Rally working nicely in VR on the motion rig. You can read more about my motion rig here: http://projectable.me/my-sim-racing-motion-rig-version-2/\n\n[![](/content/images/2017/12/IMG_3182.jpg)](http://projectable.me/my-dirt-rally-tabletop-car-cade-game/)\n\n####The problems\nThe issue I had was 2 fold: #1 the motion was too violent, and #2 even with a GTX1080Ti the framerate was really inconsistent, leading to really bad motion sickness. I had put a small amount of time in to trying to figure it out, but in the end I just gave up due to feeling quite ill.\n\n\n####The Fixes\nI tackled the motion aspect first, it took some motion profile tuning inside SimXperience Sim Commander, mainly smoothing for the device and a smoothing filter on acceleration/braking/cornering g-forces. With the heavily smoothed profile things were less back-breaking.\n\nSecondly I dived in to the graphics quality, testing how each change \"felt\" and eventually ending up with a profile that was mostly medium and low settings. The only things left on high were track and car detail. Its a bit of a shame to end up with such low settings on such a high end graphics card, but thats what I ended up with after \"gut checking\" them.\n\nAfter testing and doing a few runs I was quite happy that I no longer wanted to loose my lunch. It shows just how important high FPS without any stutter is to avoid motion sickness in VR.\n\n####Some tweaking\nDuring all the testing I found that the actuator bar was hitting the back of the seat, meaning there was a LOT of flex and disconnect between the seat and the seat platform. I quickly bent up a bracket to fix this. The seat had an M8 bolt at the top that I used to mount back to the top of the seat platform. This extra rigidity helps a LOT! It's not pretty, but it works. I had to use a rubber bushing between the bracket and the seat, as without it there was a lot of actuator noise that was resinating in the seat itself.\n\n![](/content/images/2017/12/IMG_3346-1.JPG)\n![](/content/images/2017/12/IMG_3347.JPG)\n\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/wZdebttwHCM\" frameborder=\"0\" allowfullscreen\u003e\u003c/iframe\u003e\n\n\n####Final Thoughts\nSo there we have it. Small tweaks and changes to the rig to help enhance the experience, and some game settings tweaks along with motion profile changes to make Dirt Rally a playable experience in motion VR!\n\n![](/content/images/2017/12/IMG_3348-1.JPG)\n![](/content/images/2017/12/IMG_3349.JPG)\n![](/content/images/2017/12/IMG_3350.JPG)\n\n\n####System Specifications\n\n#####PC\n- i7-5820K\n- 32G ram\n- GTX 1080Ti\n- Benq W1080ST projector projecting on the wall\n- Oculus Rift CV1\n\n#####Motion Rig\n- Custom 8020 aluminum extrusion rig\n- SimXperience DIY motion platform\n- Fanatec CSW V2\n- Fanatec CSW Pedals V3\n- Fanatec Sequential Shifter V1.5\n- Fanatec Handbrake","html":"\u003cp\u003eWith the rekindled interest in Dirt Rally that my \u003ca href=\"http://projectable.me/my-dirt-rally-tabletop-car-cade-game/\" rel=\"nofollow\"\u003eCar-cade\u003c/a\u003e has given me I had a bit of a yearning to get Dirt Rally working nicely in VR on the motion rig. You can read more about my motion rig here: \u003ca href=\"http://projectable.me/my-sim-racing-motion-rig-version-2/\" rel=\"nofollow\"\u003ehttp://projectable.me/my-sim-racing-motion-rig-version-2/\u003c/a\u003e\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"http://projectable.me/my-dirt-rally-tabletop-car-cade-game/\" rel=\"nofollow\"\u003e\u003cimg src=\"/content/images/2017/12/IMG_3182.jpg\" alt=\"\"\u003e\u003c/a\u003e\u003c/p\u003e\n\u003ch4\u003eThe problems\u003c/h4\u003e\n\u003cp\u003eThe issue I had was 2 fold: #1 the motion was too violent, and #2 even with a GTX1080Ti the framerate was really inconsistent, leading to really bad motion sickness. I had put a small amount of time in to trying to figure it out, but in the end I just gave up due to feeling quite ill.\u003c/p\u003e\n\u003ch4\u003eThe Fixes\u003c/h4\u003e\n\u003cp\u003eI tackled the motion aspect first, it took some motion profile tuning inside SimXperience Sim Commander, mainly smoothing for the device and a smoothing filter on acceleration/braking/cornering g-forces. With the heavily smoothed profile things were less back-breaking.\u003c/p\u003e\n\u003cp\u003eSecondly I dived in to the graphics quality, testing how each change \u0026#34;felt\u0026#34; and eventually ending up with a profile that was mostly medium and low settings. The only things left on high were track and car detail. Its a bit of a shame to end up with such low settings on such a high end graphics card, but thats what I ended up with after \u0026#34;gut checking\u0026#34; them.\u003c/p\u003e\n\u003cp\u003eAfter testing and doing a few runs I was quite happy that I no longer wanted to loose my lunch. It shows just how important high FPS without any stutter is to avoid motion sickness in VR.\u003c/p\u003e\n\u003ch4\u003eSome tweaking\u003c/h4\u003e\n\u003cp\u003eDuring all the testing I found that the actuator bar was hitting the back of the seat, meaning there was a LOT of flex and disconnect between the seat and the seat platform. I quickly bent up a bracket to fix this. The seat had an M8 bolt at the top that I used to mount back to the top of the seat platform. This extra rigidity helps a LOT! It\u0026#39;s not pretty, but it works. I had to use a rubber bushing between the bracket and the seat, as without it there was a lot of actuator noise that was resinating in the seat itself.\u003c/p\u003e\n\u003cimg src=\"/content/images/2017/12/IMG_3346-1.JPG\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2017/12/IMG_3347.JPG\" alt=\"\"\u003e\n\u003ciframe width=\"560\" height=\"315\" src=\"https://www.youtube.com/embed/wZdebttwHCM\" frameborder=\"0\" allowfullscreen=\"\"\u003e\u003c/iframe\u003e\n\u003ch4\u003eFinal Thoughts\u003c/h4\u003e\n\u003cp\u003eSo there we have it. Small tweaks and changes to the rig to help enhance the experience, and some game settings tweaks along with motion profile changes to make Dirt Rally a playable experience in motion VR!\u003c/p\u003e\n\u003cimg src=\"/content/images/2017/12/IMG_3348-1.JPG\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2017/12/IMG_3349.JPG\" alt=\"\"\u003e\n\u003cimg src=\"/content/images/2017/12/IMG_3350.JPG\" alt=\"\"\u003e\n\u003ch4\u003eSystem Specifications\u003c/h4\u003e\n\u003ch5\u003ePC\u003c/h5\u003e\n\u003cul\u003e\n\u003cli\u003ei7-5820K\u003c/li\u003e\n\u003cli\u003e32G ram\u003c/li\u003e\n\u003cli\u003eGTX 1080Ti\u003c/li\u003e\n\u003cli\u003eBenq W1080ST projector projecting on the wall\u003c/li\u003e\n\u003cli\u003eOculus Rift CV1\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch5\u003eMotion Rig\u003c/h5\u003e\n\u003cul\u003e\n\u003cli\u003eCustom 8020 aluminum extrusion rig\u003c/li\u003e\n\u003cli\u003eSimXperience DIY motion platform\u003c/li\u003e\n\u003cli\u003eFanatec CSW V2\u003c/li\u003e\n\u003cli\u003eFanatec CSW Pedals V3\u003c/li\u003e\n\u003cli\u003eFanatec Sequential Shifter V1.5\u003c/li\u003e\n\u003cli\u003eFanatec Handbrake\u003c/li\u003e\n\u003c/ul\u003e\n","excerpt":"","metaTitle":"","metaDescription":"","status":"published","featured":false,"publishedAt":"2017-12-03T02:51:45Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"5dfef4da0e8654d6e59930c9093f63ff","projectId":"","tags":[{"id":"ecaaace3f367d1a9865ae0933d53d715","name":"Sim Racing","slug":"sim-racing"},{"id":"9c4230a56ffd914aade79798824d9d23","name":"SimXperience","slug":"simxperience"},{"id":"ab4bc19e35909615ae7bc6a344a2b50a","name":"Motion Simulator","slug":"motion-simulator"},{"id":"bda7da089cef0417916d7e03950af340","name":"Dirt Rally","slug":"dirt-rally"}],"createdAt":"2026-10-18T13:16:27Z","updatedAt":"2026-10-18T13:16:27Z"},{"id":"94b2de371382ec793a64ed49343c14c6","siteId":"projectable","slug":"talkiepi-nov2017-updates","title":"talkiepi Updates (November 2017)","markdown":"##Introducing [talkiepi](https://github.com/dchote/talkiepi)\n\n[talkiepi](https://github.com/dchote/talkiepi) is a wifi \"walkie\" talkie for your kids and their friends. It provides a very simple \"push to talk\" interface. When you push the button and talk, all the other talkiepis in the channel will hear what is being said.\n\n##November 2017 Updates\n[talkiepi](https://github.com/dchote/talkiepi) has been out in the world for a little over a year now and it has been quite a successful project! I have seen 4+ different magazine articles (read some: [here](https://www.raspberrypi.org/magpi-issues/MagPi62.pdf), [here](https://issuu.com/futurepublishing/docs/piu02.issuu/2?ff=true), and [here](https://magazinesubscriptionsdigital.zinio.com/browse/issues/index.jsp;jsessionid=FCDE8D752BEF8FE767D9CF82D567CB65.prd-main-news2?skuId=416414901\u0026prnt=\u0026categoryId=cat1960026)), and I have even seen it included in school course curriculum!!\n\nI am getting back in to [Go](https://golang.org) development for work, and the process of working in [Go](https://golang.org) again gave me the hankering to update [talkiepi](https://github.com/dchote/talkiepi) and try and work out some of its kinks. I'll be working on enhancements to talkiepi over the next few weeks when I get the time. My goal is to get talkiepi working flawlessly on a [Raspberry Pi Zero W](https://www.adafruit.com/product/3400).\n\n**Talkiepi can be found over on github: https://github.com/dchote/talkiepi**","html":"\u003ch2\u003eIntroducing \u003ca href=\"https://github.com/dchote/talkiepi\" rel=\"nofollow\"\u003etalkiepi\u003c/a\u003e\u003c/h2\u003e\n\u003cp\u003e\u003ca href=\"https://github.com/dchote/talkiepi\" rel=\"nofollow\"\u003etalkiepi\u003c/a\u003e is a wifi \u0026#34;walkie\u0026#34; talkie for your kids and their friends. It provides a very simple \u0026#34;push to talk\u0026#34; interface. When you push the button and talk, all the other talkiepis in the channel will hear what is being said.\u003c/p\u003e\n\u003ch2\u003eNovember 2017 Updates\u003c/h2\u003e\n\u003cp\u003e\u003ca href=\"https://github.com/dchote/talkiepi\" rel=\"nofollow\"\u003etalkiepi\u003c/a\u003e has been out in the world for a little over a year now and it has been quite a successful project! I have seen 4+ different magazine articles (read some: \u003ca href=\"https://www.raspberrypi.org/magpi-issues/MagPi62.pdf\" rel=\"nofollow\"\u003ehere\u003c/a\u003e, \u003ca href=\"https://issuu.com/futurepublishing/docs/piu02.issuu/2?ff=true\" rel=\"nofollow\"\u003ehere\u003c/a\u003e, and \u003ca href=\"https://magazinesubscriptionsdigital.zinio.com/browse/issues/index.jsp;jsessionid=FCDE8D752BEF8FE767D9CF82D567CB65.prd-main-news2?skuId=416414901\u0026amp;prnt=\u0026amp;categoryId=cat1960026\" rel=\"nofollow\"\u003ehere\u003c/a\u003e), and I have even seen it included in school course curriculum!!\u003c/p\u003e\n\u003cp\u003eI am getting back in to \u003ca href=\"https://golang.org\" rel=\"nofollow\"\u003eGo\u003c/a\u003e development for work, and the process of working in \u003ca href=\"https://golang.org\" rel=\"nofollow\"\u003eGo\u003c/a\u003e again gave me the hankering to update \u003ca href=\"https://github.com/dchote/talkiepi\" rel=\"nofollow\"\u003etalkiepi\u003c/a\u003e and try and work out some of its kinks. I\u0026#39;ll be working on enhancements to talkiepi over the next few weeks when I get the time. My goal is to get talkiepi working flawlessly on a \u003ca href=\"https://www.adafruit.com/product/3400\" rel=\"nofollow\"\u003eRaspberry Pi Zero W\u003c/a\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTalkiepi can be found over on github: \u003ca href=\"https://github.com/dchote/talkiepi\" rel=\"nofollow\"\u003ehttps://github.com/dchote/talkiepi\u003c/a\u003e\u003c/strong\u003e\u003c/p\u003e\n","excerpt":"","metaTitle":"","metaDescription":"","status":"published","featured":false,"publishedAt":"2017-11-17T15:15:46Z","authorId":"190160de54f2192bd5b8734ec8808086","authorName":"Daniel Chote","authorSlug":"daniel","heroMediaId":"f1b2883a1a2e99060f5e4a6a2f8ccebd","projectId":"","tags":[{"id":"b223b3e2ba0e78e7e6a6a8ff4349517d","name":"talkiepi","slug":"talkiepi"},{"id":"b529314e428273696b02c474dcf56132","name":"Raspberry Pi","slug":"raspberry-pi"},{"id":"88aab2a2ba262f200b248f0074011445","name":"WIFI","slug":"wifi"},{"id":"e2e48067ff28749272dbdc37a0f2023d","name":"golang","slug":"golang"},{"id":"8600a98484e169ed96ea6fbe2a034ccc","name":"VOIP","slug":"voip"}],"createdAt":"2026-10-18T13:16:27Z","updatedAt":"2026-10-18T13:16:27Z"}],"site":{"id":"projectable","theme":"projectable","title":"Daniel Chote's Project Blog","description":"Software Development, Game Development, Hardware Hacking, Sim Racing, Multi-rotors and Robots!","timezone":"America/New_York","postsPerPage":10,"logoMediaId":"21557257b3f2ecc4b3fa875b808097a7","coverMediaId":"6d7e99f0997d13aac98da38677ca7064","copyright":"© Daniel Chote","twitterUrl":"https://twitter.com/dchote","linkedinUrl":"https://www.linkedin.com/in/dchote","googleTagId":"G-07GJBQVM38","hosts":[{"host":"new.projectable.me","canonical":false},{"host":"projectable.me","canonical":true},{"host":"www.projectable.me","canonical":false}]},"tagId":"","total":157}

Daniel Chote's Project Blog

Software Development, Game Development, Hardware Hacking, Sim Racing, Multi-rotors and Robots!

Shields up: new ships, Pulse on the grid, and a look inside NullG
A busy week of rebuilt XL ships, learned pilots, shields and split-screen polish, with Pulse on the grid and a closer look at the MCP tools behind it.
Recreating the WipEout XL fleet in Blender
A closer look at NullG’s recreated XL ships and trackside models, from constrained cockpit glass and worn paint to Blender MCP, animated previews, and
A new soundtrack for NullG
Have a listen to NullG’s new soundtrack: seven Suno-made instrumental tracks inspired by 90s drum and bass, jungle and EDM.
Teaching NullG to race: the reinforcement-learning journey so far
From native driving controls and road guidance to Turbo preservation and weapon combat: the experiments, references and repeated adjustments behind NullG’s
Introducing NullG: anti-gravity racing and AI-assisted development
NullG’s development from January’s engine plans through asset loading, racing systems, XL research, agent validation, and reinforcement learning.
Ubuntu 24.04 + Nvidia Drivers + Ollama
I am building an updated server to play with LLMs for local home automation control using Home Assistant ( https://www.home-assistant.io ). I managed to pick
Comcast Business IPv6 on Ubiquiti Unifi
If you are reading this it is likely you were like me, on the edge of giving up on IPv6 on your network. After combining a few small pieces from posts
A simple Etcher 6 slot SD card jig
I've been working with Kubernetes a lot lately, I am using Kubernetes at Filament as a way to manage sandbox blockchain environments, so naturally I wanted to
Sim racing motion rig v2 updates
With the rekindled interest in Dirt Rally that my Car-cade has given me I had a bit of a yearning to get Dirt Rally working nicely in VR on the motion rig. You
talkiepi Updates (November 2017)
Introducing talkiepi talkiepi is a wifi "walkie" talkie for your kids and their friends. It provides a very simple "push to talk" interface. When you push the