Axis Robotics

Axis is scaling into a model company. We built the data infrastructure for one reason: to produce general robotic intelligence. Our approach is centered on skill: we turn reinforcement learning experts into reusable, conditioned skills,…

We tested @dreamlabs_ai's API on our own YAM Box from @i2rt_robotics with a natural language instruction ⬇️ We're working toward a similar vision, models that generalize across tasks and embodiments, and this is a promising glimpse of…

The final snapshot is locked in. Season 1 wraps up today. A huge thank you to every contributor who's been building with us. Season 2 kicks off on Oct 23, so there won't be any tasks from Oct 18 through Oct 22. The last batch of Points,…

Thanks @KhalaResearch for featuring us in their latest report. It covers where robotics stands today and where the hard problems are, before taking a close look at Axis: strategic positioning, product matrix, academic achievements,…

This demo shows reinforcement learning running the full control loop for a dual-arm robot: picking up a kettle and pouring water. It highlights how much RL can bring to dynamics adaptation, disturbance recovery, and complex contact…

Physical AI requires real plant floors to keep learning. Factory floors build the compounding data loop. The flywheel only accelerates when systems learn from continuous operational experience. Physical experience turns raw demos into…

Simulation provides scalable expertise. Vision distillation brings it to reality. We are training RL experts in simulation and distilling them into vision policies that generalize. Here's a pick-and-place expert that handles objects placed…

A drawing describes the nominal part. The robot welds the actual one. When an operator points a tablet camera at an engineering drawing, generating robot motion is only half the job. The real part was cut, bent, and tacked within…

A robotics trajectory is not truly complete just because the task is finished. @axisrobotics makes the final verification step important. Contributors record robot movements in browser based simulation tasks, submit their trajectories for…

Stage demos show isolated movements, but factory floors measure uptime and parts variance. Handling hundreds of part geometries without custom tooling proves the real value of synthetic training. https://t.co/PoR7ku1Goi

The shift from collecting robotics datasets to executing factory floor sorting is where real value gets created. High-mix automotive lines cannot afford bespoke engineering for every bracket and fastener. https://t.co/LEfWxUP7mU

Handling 500 to 800 distinct automotive components under one simulation-trained policy changes the unit economics of automation. Zero manual intermediate engineering is the benchmark production lines actually care about.…

Continuous learning loops in robotics matter far more than static datasets. Moving from task generation in simulation to live parts sorting at Lotus and Geely validates the data engine on the plant floor. https://t.co/YJjCqmhtJ7

Scaling parts sorting across a production line requires moving beyond demos. Axis Robotics deploying on Lotus and Geely lines takes Sim2Real straight into live manufacturing. Sim2Real provides the scalable pipeline needed for high-mix…

Announcing our commercial partnership with @lotuscars @GeelyAutoGlobal Lotus Cars builds iconic high-performance sports cars and hypercars. Through our Axis Suite, we deliver an end-to-end Sim2Real automation solution for production-line…

What a night at @wallet! We brought two K1 robots from @boosterobotics, one of our key embodiment partners. They put on a show and interacted with everyone at the venue, showcasing the robotic intelligence we're building. Thanks to…

What a night at @wallet. Thanks to everyone who stopped by, waved, and said hi to our little robots. They absolutely loved every bit of it, and we hope they brought a few smiles. See you next time! https://t.co/tjN6Yr9LHe…

The Value of Simulation Data for Real-World Robot Learning - An Empirical Study

Agreed. The sim-to-real gap on contact dynamics is the bottleneck I keep coming back to. A robot can plan a trajectory in simulation all day. The moment it touches a flexible harness or seats a clip on a moving line, minor errors compound…

Task allocation across manipulators is the upstream constraint. On an automotive line, assigning the wrong arm to a clip insertion compounds into seconds of takt loss per cycle. The motion plan is downstream of that decision.…

Thanks @baseapac for featuring us. September was huge, and October will bring more! https://t.co/VjIt9GlQuK

KBW @kbwofficial was a hit. In Seoul we got to meet contributors from all over the world, and talk through the Physical AI data engine we're building. Thanks to everyone who stopped by and built with us. See you all next time.…

Full text pending

Announcing our partnership with @ManycoreTech Manycore Tech is a leading global provider of spatial intelligence, focused on accelerating the integration of AI into the physical world. Axis Suite delivered a set of complex, physics-ready…

Watch our CEO @chris_anm01 and @brian_armstrong share how they see robotics changing the world ⬇️ https://t.co/4PRhq9yNk7

This demo shows a 97% success expert policy trained on a single RTX 4090 in ~20 minutes. The policy starts with no idea how to do the task and ends up completing it fluently, rollout after rollout. Read the blog here:…

Physical AI's GPT moment is closer than you think. Our CEO @chris_anm01 sat down with @brian_armstrong to explain why, and how Axis is building towards it on @base. Watch ⬇️ https://t.co/bA6iqsWbio

A new scaling paradigm for robotic model training: from raw trajectories to reusable, composable and distillable skills. Read the full blog: https://t.co/fHQHia3ULZ Three findings from our latest experiments: • Embodied RSI: the deployed…

Introducing the Open Axis Benchmark, a living benchmark engine for honest evaluation of robot manipulation models, built together with @openroboto. Robotic models evolve faster every month, while most benchmarks stay frozen. Models overfit…

We're bringing robots to the rooftop with @wallet. See you there 🦾 https://t.co/K0S8a5X9GN

Thrilled our paper has been accepted to the Physical World Models (PWMS) workshop at #IROS2026 🎉 Growable, community-driven data engine for robot manipulation: 1,800+ tasks, 1.5M+ trajs. π0.5+AXIS: 83.9→88.8 on LIBERO-Plus. See you in…

Announcing our commercial partnership with @FeagineRobotics Feagine builds soft embodied intelligence infrastructure—from bionic flexible robot arms to embodiment-free data and the Fi0 cross-embodiment foundation model. Through our…

We just crossed 5M robot trajectories on @base, generated by 200K+ contributors worldwide. LLMs had the internet. We're building its physical-experience equivalent for robots. The future of robotics is being built onchain, on @base.…

Our CEO sat down with @brian_armstrong in Singapore last week to show what we're building. Data is Physical AI's biggest bottleneck, and it's exactly what our data engine is built to solve. Scaling it this fast wouldn't be possible without…

Physical AI will be built on data — by everyone, everywhere. Thanks @BasedIndia for having us. Watch how we're accelerating next-gen robotics ⬇️ https://t.co/HDJESNH4X9

Get to know Axis Robotics in under 2 minutes ⬇️ https://t.co/69lBbL6DNb

Axis Weekly We've been experimenting with how to digest data better. After almost a year of building the data flywheel, the next step is a model-guided data engine: use foundation-model performance as feedback to decide what to keep, what…

Compute deals in the billions, a $1.6B perception raise, and new ways to collect robot data without even using the robot. 1️⃣ @Figure_robot partnered with Nscale to deploy up to 100,000 GPUs on NVIDIA Vera Rubin, with $3.5B committed…

Post Log in Sign up Post Omai Leidi (3/3) on X: "Happy Wednesday frens👋 Everyone is watching the race to build better robots. I’m starting to think the real race is the data teaching those robots how to act. I’ve been digging into…

Live campaign Signal Boost Campaign Connect your X account and follow @axisrobotics — your badge lights up as soon as we see the follow. Connect →+1 PTS · credited instantly Total Trajectories 4,794,179 Total Users 196,995 7D New Tasks 494…

i thought community robot data would hit diminishing returns fast but Axis Dataset V1 just made that take look stupid @axisrobotics pushed π0.5 from 83.9% to 88.8% on LIBERO Plus, and performance kept climbing as training data scaled from…

Live campaign Signal Boost Campaign Connect your X account and follow @axisrobotics — your badge lights up as soon as we see the follow. Connect →+1 PTS · credited instantly Total Trajectories 4,794,218 Total Users 197,003 7D New Tasks 494…

Post Log in Sign up Post - Cicada Market Making @cicada\mm Inside Robotics and Physical AI Featuring insights from Axis Robotics In 2026, the artificial intelligence industry learned to solve the compute problem. GPUs are becoming more…

Announcing our partnership with @Dexmal_AI About Dexmal: Dexmal is a technology company focused on general-purpose embodied intelligence, committed to building intelligent, useful, and trustworthy robots. As a core data infrastructure…

Many People believe Physical might be unachievable not because robots lack arms or GPUs rather, because robots do not have enough real training data. Normal AI LLM can train on text that already exists on the internet but Robots cannot. A…

Back How task-aligned simulation, real demonstrations, and continued pretraining can make each new Booster task easier to build than the last. Robot learning has a model problem. More importantly, it has a data-operations problem. Figure —…

Back Axis Robotics launches an exclusive robot-data campaign on Binance Wallet. Binance Keyless Wallet users get their own task set on Axis Hub and their own reward pool of 1,500,000 Axis Points, separate from the regular product points…

2/ What you're actually doing Axis is the compounding data engine for physical AI. Every task you complete on Axis Hub is a robot trajectory — real training data for real robot policies. You do the demonstration, we verify it, you sign it…

Post Log in Sign up Post Axis Robotics on X: "Axis Robotics x @BinanceWallet is live. 1,500,000 Axis Points. 30 days. Reserved exclusively for Binance Keyless Wallet users. Teach robots. Sign your data on Base. Get paid in Points from a…

Axis Weekly Last week, we closed the remaining replay and runtime gaps between the browser, policy server, and physics stack — then scaled the coverage of our task generation engine (TaskGen). An articulated asset library is now in the…

Back Introduction DAgger-style correction is the engine of post-training improvement: when a policy fails, a human takes over, steers the robot back on track, and the resulting trajectory goes back into training. Since launching Axis V2,…

01Getting Started16 What do I need to use Axis Hub? A modern browser (Chrome, Firefox, Safari, or Edge) on a desktop or laptop. No GPU, no downloads, no special hardware. The physics engine — MuJoCo, compiled to WebAssembly — runs locally…

Post Log in Sign up Post Axis Robotics on X: "Finally, we’ve made it into the Top 3 DApps on @Base. Base is the hub for Physical AI. 🤖" - Axis Robotics @axisrobotics Finally, we’ve made it into the Top 3 DApps on @base. Base is the hub…