https://roughcut.dev/Notes on physical AI, simulation, digital twins, industrial systems, and developer tools.en-usSat, 29 Aug 2026 00:00:00 GMThttps://roughcut.dev/writing/most-robot-tasks-dont-need-touchhttps://roughcut.dev/writing/most-robot-tasks-dont-need-touchThu, 27 Aug 2026 00:00:00 GMTYears of MES work taught me where automation stops: at tasks whose decisive state cannot be written down or seen. Tactile input earns its place when it reveals that missing state; contact alone is not enough.Physical AIIndustrial Systemshttps://roughcut.dev/projects/i-took-the-order-away-from-the-agent-then-learned-order-wasnt-enoughhttps://roughcut.dev/projects/i-took-the-order-away-from-the-agent-then-learned-order-wasnt-enoughSat, 22 Aug 2026 00:00:00 GMTAnchor stopped my agent from skipping declared steps. Ratchet came from the next question: after an interrupted or overlapping run, could I prove what happened and restart without making it worse?Agentic Systemshttps://roughcut.dev/projects/building-ratchet-runtimehttps://roughcut.dev/projects/building-ratchet-runtimeSat, 22 Aug 2026 00:00:00 GMTA technical look at the small execution runtime I built for overlapping and interrupted agent workflows: ownership, effect recovery, verified completion and the limits of a local alpha.Agentic Systemshttps://roughcut.dev/projects/a-gaussian-splat-is-not-yet-a-digital-twinhttps://roughcut.dev/projects/a-gaussian-splat-is-not-yet-a-digital-twinFri, 21 Aug 2026 00:00:00 GMTA Gaussian splat can reproduce a place convincingly while knowing almost nothing about its identity, scale or behavior. This article separates reconstruction, visual readiness, operational readiness and simulation readiness—and defines the boundary SplatStage is designed to cross.Simulation & Digital Twinshttps://roughcut.dev/projects/building-splatstage-from-edited-gaussians-to-openusdhttps://roughcut.dev/projects/building-splatstage-from-edited-gaussians-to-openusdFri, 21 Aug 2026 00:00:00 GMTSplatStage turns a Gaussian reconstruction into a versioned editing workflow, exports the selected PLY as an OpenUSD Gaussian ParticleField, and composes it with replaceable engineered assets. The hardest bug was at the seam between editing and export.Simulation & Digital Twinshttps://roughcut.dev/projects/what-5-8-million-gaussians-taught-me-about-building-a-visual-twinhttps://roughcut.dev/projects/what-5-8-million-gaussians-taught-me-about-building-a-visual-twinFri, 21 Aug 2026 00:00:00 GMTOne Garden scene made the promises and gaps in SplatStage measurable: 2.07 million Gaussians removed, an edited OpenUSD particle field, three composed assets, a dependency-complete stage—and no measured scale, colliders or external runtime proof.Simulation & Digital Twinshttps://roughcut.dev/writing/what-kind-of-truth-does-your-digital-twin-needhttps://roughcut.dev/writing/what-kind-of-truth-does-your-digital-twin-needSat, 15 Aug 2026 00:00:00 GMTEvery digital twin needs a usable digital starting point. CAD, LiDAR, photogrammetry and Gaussian splatting preserve different kinds of truth, so the representation should follow the decision. Part 1 of a three-part series.Simulation & Digital TwinsIndustrial Systemshttps://roughcut.dev/projects/from-camera-media-to-gaussian-splats-building-reconstudiohttps://roughcut.dev/projects/from-camera-media-to-gaussian-splats-building-reconstudioSat, 15 Aug 2026 00:00:00 GMTReconStudio turns camera media into Gaussian splats and OpenUSD through a browser and job API. I built it to expose reconstruction stages, retain experiment evidence and test where on-demand GPU execution fits. Part 2 of a three-part series.Simulation & Digital Twinshttps://roughcut.dev/projects/what-i-learned-from-24-gaussian-splat-jobshttps://roughcut.dev/projects/what-i-learned-from-24-gaussian-splat-jobsSat, 15 Aug 2026 00:00:00 GMTI built ReconStudio, then used 24 job records across six scenes to test its evidence. Metric bugs, repeated runs and paired-frame comparisons changed what I could claim about training, capture and GPU capacity. Part 3 of a three-part series.Simulation & Digital Twinshttps://roughcut.dev/writing/what-industrial-maintenance-demands-from-robot-capturehttps://roughcut.dev/writing/what-industrial-maintenance-demands-from-robot-captureSat, 08 Aug 2026 00:00:00 GMTA refinery valve operation shows why robot capture should begin with the task: preserve the motion, contact, timing and measurements the policy will need.Physical AIIndustrial Systemshttps://roughcut.dev/writing/a-vla-benchmark-is-not-a-commissioning-packagehttps://roughcut.dev/writing/a-vla-benchmark-is-not-a-commissioning-packageSun, 02 Aug 2026 00:00:00 GMTRobot foundation models are getting better at handling situations they were not shown. That makes the old commissioning questions more important: where is the model valid, how do I know it has left that range, and what happens next?Physical AIIndustrial Systemshttps://roughcut.dev/links#foresightsafety-vlahttps://roughcut.dev/links#foresightsafety-vlaSat, 01 Aug 2026 00:00:00 GMTA robot can finish the task and still behave unsafely along the way. That distinction makes this benchmark worth reading.Physical AIIndustrial Systemshttps://roughcut.dev/projects/my-agent-kept-skipping-stepshttps://roughcut.dev/projects/my-agent-kept-skipping-stepsTue, 21 Jul 2026 00:00:00 GMTA repeatable agent workflow kept skipping different steps on different runs. The fix wasn't a stronger prompt; it was moving the plan into a DAG and letting deterministic code control sequence and verification.Agentic Systemshttps://roughcut.dev/writing/nvidia-hardware-is-one-architecture-in-many-boxeshttps://roughcut.dev/writing/nvidia-hardware-is-one-architecture-in-many-boxesSun, 19 Jul 2026 00:00:00 GMTPeople keep asking me a version of the same question: what NVIDIA hardware should I be thinking about for a digital twin, from a developer build at my desk all the way to production? I always had an answer but no clean model behind it, just a pile of product names, three of which are confusingly all called DGX. So I built the model I wanted: NVIDIA develops one architecture family a year and expresses it across many kinds of computer, from a gaming card to a liquid-cooled rack. These are my rough notes, in case you're coming at it from the software side and asking the same thing.Physical AISimulation & Digital Twinshttps://roughcut.dev/projects/testing-the-robot-training-loophttps://roughcut.dev/projects/testing-the-robot-training-loopSat, 11 Jul 2026 00:00:00 GMTA month ago I drew a five-stage loop for a robot training center and admitted most of it was an educated guess I hadn't tested. So I built the loop as a real, orchestrated pipeline on one GPU and carried two tasks around it — a pole that balances by trial and error, and a Franka arm that learns to stack cubes by copying demonstrations. This is the environment, the stack, the architecture, and what actually ran. Part 1 of two.Physical AISimulation & Digital Twinshttps://roughcut.dev/projects/what-the-robot-training-loop-taught-mehttps://roughcut.dev/projects/what-the-robot-training-loop-taught-meSat, 11 Jul 2026 00:00:00 GMTRunning the training loop for real surfaced one concept the diagram doesn't capture: the same five stages take two different shapes depending on how the robot learns — by trial and error (reinforcement learning) or by copying demonstrations (imitation). This part is the mechanism behind that fork, grounded in the two tasks I ran, plus the open questions the build left unproven. Part 2 of two.Physical AIhttps://roughcut.dev/writing/most-teams-reach-for-fine-tuning-too-earlyhttps://roughcut.dev/writing/most-teams-reach-for-fine-tuning-too-earlyWed, 08 Jul 2026 00:00:00 GMTWhen a general-purpose LLM (or an agent built on it) doesn't fit the problem, the reflex is to fine-tune it. I think that's usually the third thing to try, not the first. There are three ways to close the gap: grounding context, harness optimization, and fine-tuning, and which one you reach for should depend on the need in front of you: domain knowledge, agent behavior, or raw performance.Agentic Systemshttps://roughcut.dev/writing/is-physical-ai-the-next-level-for-meshttps://roughcut.dev/writing/is-physical-ai-the-next-level-for-mesTue, 07 Jul 2026 00:00:00 GMTMES was never really about running the machines. It was about improving the operation, and it was limited by how much had to be modeled by hand and decided by people. Physical and agentic AI can change that, but only if the plant's systems are joined through shared operational context.Industrial SystemsAgentic Systemshttps://roughcut.dev/links#world-models-capital-shifthttps://roughcut.dev/links#world-models-capital-shiftFri, 03 Jul 2026 00:00:00 GMTThe useful signal is not that world models are solved. It is that investors and research leaders increasingly see predictive models of physical dynamics as a distinct bottleneck worth funding.Physical AIhttps://roughcut.dev/projects/live-usd-viewerhttps://roughcut.dev/projects/live-usd-viewerThu, 02 Jul 2026 00:00:00 GMTA browser-based USD viewer, RTX-rendered server-side and streamed over WebRTC, built to see if the library-first pitch holds up under real interactivity.Simulation & Digital Twinshttps://roughcut.dev/writing/physical-ai-bottleneck-isnt-the-modelhttps://roughcut.dev/writing/physical-ai-bottleneck-isnt-the-modelWed, 01 Jul 2026 00:00:00 GMTWe finally have models that can watch the physical world, reason about it, and predict what happens next. Walk a real automotive parts-kitting task through what they actually solve, and the thing still standing between a demo and a working line isn't a smarter model: it's the policy you have to train for your machine, and the system that keeps it trained.Physical AIIndustrial Systemshttps://roughcut.dev/writing/no-longer-need-an-apphttps://roughcut.dev/writing/no-longer-need-an-appWed, 24 Jun 2026 00:00:00 GMTKit and libraries used to be the same choice for building on Omniverse. They're not anymore, and that's a bigger change than it looks, with a real before/after from a Kit extension I built.Simulation & Digital Twinshttps://roughcut.dev/writing/the-omniverse-license-change-you-might-have-missedhttps://roughcut.dev/writing/the-omniverse-license-change-you-might-have-missedWed, 17 Jun 2026 00:00:00 GMTSince May 2026, Omniverse is free for production, not just development. It's a clean, sensible change, but it landed quietly, and I keep meeting customers and partners who don't know it happened yet. The paid-production requirement is gone; the license and optional enterprise support remain.Simulation & Digital Twinshttps://roughcut.dev/writing/a-robot-training-center-is-one-loophttps://roughcut.dev/writing/a-robot-training-center-is-one-loopMon, 15 Jun 2026 00:00:00 GMTSomeone asked me how you'd build a training center for industrial robots from scratch, and my first reaction was that it isn't one project, it's ten. What untangled it was seeing that they aren't ten separate things at all. They're five stages of one loop that turns real work into robot skills and feeds what happens back in. This is where I've gotten so far, still learning, sharing it in case you're walking the same road.Physical AIhttps://roughcut.dev/writing/omniverse-is-not-a-rendering-toolhttps://roughcut.dev/writing/omniverse-is-not-a-rendering-toolWed, 03 Jun 2026 00:00:00 GMTFrom the industrial systems I know, rendering is only one part of Omniverse. The harder job is connecting source data, giving teams a shared context, and reusing it for visibility, simulation, optimization and Physical AI.Simulation & Digital TwinsIndustrial Systems