--- title: Antioch Robotics wants CI/CD for robots, and the $32M... description: Antioch Robotics raised a $32M Series A to build CI/CD-style release gates for physical AI, using simulation (Isaac Sim/Isaac Lab) to reproduce and vary... canonical: https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Antioch Robotics wants CI/CD for robots, and the $32M bet makes sense | daily.dev og:description: Antioch Robotics raised a $32M Series A to build CI/CD-style release gates for physical AI, using simulation (Isaac Sim/Isaac Lab) to reproduce and vary... og:url: https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz og:image: https://api.daily.dev/og/posts/iZywMcSQZ.png og:image:alt: Antioch Robotics wants CI/CD for robots, and the $32M bet makes sense og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Antioch Robotics wants CI/CD for robots, and the $32M bet makes sense **[Trends](https://daily.dev/sources/trends)** · 2 min read · 1 upvotes · 0 comments ## Summary Antioch Robotics raised a $32M Series A to build CI/CD-style release gates for physical AI, using simulation (Isaac Sim/Isaac Lab) to reproduce and vary real-world robot failures as repeatable tests fanned out across cloud GPUs. An agent-driven loop automates building scenes, running robots, and iterating on failures, aiming to bring software-style release cycles to robotics, betting that simulation can finally be trustworthy enough at the right fidelity to catch nasty, condition-specific bugs before deployment. ## Content Anyone who's shipped code for physical hardware knows the pain: you can't just run a test suite. You have to run the actual robot. Santiago Pino, who worked on computer vision for robots and drones, put it bluntly: "nothing is slower than having to run the robot in order to test your code." Antioch is betting $32m (Series A led by Greylock, total $40.5m) that this doesn't have to be true. The pitch is straightforward: robotics teams should be able to test a new release the way software teams run CI/CD. Write a test in Isaac Sim or Isaac Lab, define your parameter distributions, fan it out across thousands of parallel cloud replicas, and gate your release on the results. No GPU required locally. No physical hardware required at all. The technical stack is more interesting than the funding announcement. Antioch ships 140+ robot models, 86 sensor models with real noise and dropout behavior, and 12 swappable physics solvers. The simulations are calibrated against real-world data, so you're not just running in a clean virtual sandbox - you're running against a digital twin that reflects how your actual hardware behaves. When a robot fails in deployment, you can reproduce that exact failure in simulation, vary the conditions around it, and keep that test forever. The harder problem they're quietly solving: in robotics, bugs don't come from bad logic alone. They come from weird combinations of wind, lighting, sensor noise, geometry, and human traffic all hitting at once. Turning that kind of physical state into a repeatable, version-controlled test object is genuinely hard. That's the gap Antioch is trying to close. An agent layer handles the tedious parts - onboarding a robot from a CAD file, designing the scenario suite, iterating on failures - all inside a browser. The whole test loop stays closed without touching hardware. Amazon's Ring and Launchpad Build AI are already customers. The round fits a broader wave of physical AI tooling investment, alongside a16z's $1.1bn robotics fund and a cluster of robot-training-data startups. The real question isn't whether simulation-as-release-gate is a good idea. It obviously is. The question is whether the sim-to-real gap is narrow enough yet that teams will actually trust it. Antioch's answer is: calibrate against real data, and it gets close enough to matter. ## Questions this post answers ### How can I test robot or drone computer vision code without running the physical hardware? Antioch's platform runs NVIDIA Isaac Sim and Isaac Lab in the cloud, letting teams write a test, define conditions, and fan it out across cloud GPUs as thousands of parallel simulations without needing a local GPU or physical robot. It includes a library of 140+ robot models, 86 sensor models with realistic noise and dropout, and 12 swappable physics solvers. _daily.dev surfaces platforms like this for teams weighing simulation-based testing against physical hardware runs._ ### Why is reproducing robotics bugs in the physical world so difficult? Robotics bugs often depend on rare combinations of conditions, such as a specific wind speed, a lighting edge case, sensor dropout at a precise moment, or unexpected contact dynamics, which are nearly impossible to reliably reproduce on real hardware. Simulation environments let teams capture the exact failure state and vary conditions around it, then lock that scenario into a regression test for every future release. _developers hardening physical AI pipelines can follow simulation-as-release-gate approaches on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 4 discussions and 32 comments across x (as of 2026-09-10). **TL;DR:** Reaction is largely enthusiastic, with developers framing the core value as breaking the slow, hardware-bound iteration loop, while a smaller thread of voices flags that the sim-to-real gap and physics fidelity remain the real hard problem. **Sentiment:** 68% positive · 24% mixed · 8% skeptical **The case for** - Cuts out costly and slow physical hardware iteration, catching bugs before risking real robots or drones. - Simulating messy real-world sensor noise (e.g. a dirty lidar in rain) rather than clean conditions is seen as the part that actually matters. - Letting an agent decide what to test, not just run given tests, is called out as the more interesting technical leap. - Enables testing dangerous or hard-to-schedule scenarios (e.g. night flight) safely and on demand. **The pushback** - Sim-to-real gap still bites even with a good digital twin, so virtual testing doesn't fully replace real-world validation. - Building an accurate physics engine is flagged as the genuinely hard part of the whole approach. - Skepticism that the handoff from model to real machine is where problems can still hide. **By community** - x (positive): Mostly upbeat and hype-driven reactions welcoming the shift from hardware-bound testing to simulation, with a handful of more grounded technical caveats about sim-to-real fidelity. **Open questions** - How well does the simulated sensor noise and physics actually hold up against real-world edge cases at deployment? - Where exactly does the model-to-machine handoff introduce failures that simulation can't catch? **Highlights** > @testingcatalog The 86 sensor models "carrying noise and dropout" is the line that matters — simulating a clean lidar is easy, simulating a dirty one on a rainy Tuesday is where the sim-to-real gap actually lives. Robotics has been bottlenecked on iteration count for a decade, and selling > — [ricci\_nov on x](https://x.com/ricci_nov/status/2097369800641048676) > @testingcatalog 12 swappable physics solvers and the agent goes from CAD to test suite without human loop. the interesting part isn't the simulation — it's that the agent decides what to test, not just runs what you give it. > — [gunbit\_ on x](https://x.com/gunbit_/status/2097358134071165083) > @svpino This is the right problem. Robot CV didn’t stall on training. It stalled on “did this break at 30° tilt in late-afternoon light?” and having to fly the thing to find out. If the twin fools the model not just the human that loop finally looks like software. > — [LearnWithBrij on x · 1 points](https://x.com/LearnWithBrij/status/2097353632240775228) > @svpino the part that surprised me is that most of that wall clock isn't the robot running. it's resetting the scene between trials. the trial is seconds. putting the world back to a known start state is minutes, and it's manual. which is why sim feels like such a jump — not that it > — [waghweb on x · 1 points](https://x.com/waghweb/status/2097382980931269006) > @svpino Quiet dashboards still need a walk-over check; a digital twin can make experiments cheaper, but the handoff from model to machine is where I look for the lie. > — [preyforge on x](https://x.com/preyforge/status/2097390028397670732) **Source threads** - [x](https://x.com/svpino/status/2097341953172263311) · 0 points · 25 comments - [x](https://x.com/trikcode/status/2097342140485456285) · 0 points · 3 comments - [x](https://x.com/rohanpaul_ai/status/2097348553513509161) · 0 points · 1 comments - [x](https://x.com/testingcatalog/status/2097351635722740137) · 0 points · 3 comments ## Similar posts on daily.dev - [Raindrop hits $50m in funding to catch AI agents failing in production](https://daily.dev/posts/raindrop-hits-50m-in-funding-to-catch-ai-agents-failing-in-production-ss50usu5b) · The Next Web · 0 upvotes · 0 comments - [Physical AI chip developer SiMa.ai hits $1.45B valuation](https://daily.dev/posts/physical-ai-chip-developer-sima-ai-hits-1-45b-valuation-qeyb8zdjh) · TechCrunch · 0 upvotes · 0 comments - [Anthropic raises $30 billion in Series G funding at $380 billion post-money valuation](https://daily.dev/posts/anthropic-raises-30-billion-in-series-g-funding-at-380-billion-post-money-valuation-xbxblnlim) · Hacker News · 3 upvotes · 0 comments --- Tags: [#venture-capital](https://daily.dev/tags/venture-capital), [#cicd](https://daily.dev/tags/cicd), [#robotics](https://daily.dev/tags/robotics) [View this post on daily.dev](https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Antioch Robotics wants CI/CD for robots, and the $32M bet makes sense","url":"https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz"},"datePublished":"2026-09-08T15:37:53.534Z","dateModified":"2026-09-10T14:55:42.217Z","description":"Antioch Robotics raised a $32M Series A to build CI/CD-style release gates for physical AI, using simulation (Isaac Sim/Isaac Lab) to reproduce and vary...","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"venture-capital,cicd,robotics","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Antioch Robotics wants CI/CD for robots, and the $32M bet makes sense"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/antioch-robotics-wants-ci-cd-for-robots-and-the-32m-bet-makes-sense-izywmcsqz#faq","mainEntity":[{"@type":"Question","name":"How can I test robot or drone computer vision code without running the physical hardware?","acceptedAnswer":{"@type":"Answer","text":"Antioch's platform runs NVIDIA Isaac Sim and Isaac Lab in the cloud, letting teams write a test, define conditions, and fan it out across cloud GPUs as thousands of parallel simulations without needing a local GPU or physical robot. It includes a library of 140+ robot models, 86 sensor models with realistic noise and dropout, and 12 swappable physics solvers. daily.dev surfaces platforms like this for teams weighing simulation-based testing against physical hardware runs."}},{"@type":"Question","name":"Why is reproducing robotics bugs in the physical world so difficult?","acceptedAnswer":{"@type":"Answer","text":"Robotics bugs often depend on rare combinations of conditions, such as a specific wind speed, a lighting edge case, sensor dropout at a precise moment, or unexpected contact dynamics, which are nearly impossible to reliably reproduce on real hardware. Simulation environments let teams capture the exact failure state and vary conditions around it, then lock that scenario into a regression test for every future release. developers hardening physical AI pipelines can follow simulation-as-release-gate approaches on daily.dev."}}]} ```