Axis just hit #2 out of 1,098 apps on the @base popular apps leaderboard!
📊 13,482 Weekly Active Users
Through @base, Axis seamlessly brings task creation, verification, scoring, and reward reconciliation fully on-chain.
From powering our Proof of Quality Work mechanism to embedding end-to-end data provenance directly into every transaction, @base makes distributed Physical Intelligence transparent, auditable, and truly trustless.
Axis Weekly
This week, we consolidated our June progress into a clearer data-loop direction: moving beyond standard short-horizon single-arm demonstrations toward complex-task data, correction data, and continuous model iteration.
Key updates:
- Teleoperation UX: We improved direct gripper dragging, object selection, and bimanual control to skip low-information actions and preserve high-value demonstration segments.
- Data quality: We strengthened verification and checker logic against new cheating patterns, extending stricter validation to bimanual tasks and DAgger collection.
- Model iteration: The automated task-to-policy loop is now largely connected, and we are shifting toward DAgger-style correction data to better distinguish human intervention from policy rollouts.
- TaskGen: Articulated-object support expanded beyond six categories, using a coding agent for asset generation and a semantic LLM agent with DINO for better asset retrieval.
- Real-world validation: Dataset v2 long-horizon data collection is underway, with early real-world results suggesting AXIS + DROID co-training preserves useful learned priors.
Details below 🧵
Alright, it’s time for another giveaway.
i’ll be airdropping $1,000 to 5 random winners.
all you need to do:
• follow
• like
• retweet this post
in 24 hours, grok will randomly pick the 5 winners.
Good luck everyone.
Hey @grok, 24 hours after this post, randomly pick 5 lucky people from the comments who completed all the requirements above for me.
Axis Weekly
This week we focused on making browser-based robot control smoother, preparing longer-horizon articulated-object tasks, and tightening the path from task data to trained and evaluated policies.
The main theme was improving the robotics loop end to end: interaction, replay, verification, training, and evaluation all moved toward workflows that are easier to inspect, reproduce, and explain.
Key updates:
- Teleoperation: End-effector dragging was rebuilt with smoother joint-space interpolation, reducing arrival jitter and raising the effective control frequency while dragging.
- Long-horizon tasks: The task set expanded toward multi-step articulated-object demos, including tasks with hinged or movable objects and four to five meaningful action steps.
- Policy release loop: The policy workflow became more explicit by connecting task IDs, model paths, evaluation outputs, success-rate summaries, and visual diagnostics.
- Multi-embodiment dataset: We validated dataset generation across multiple robot embodiments, moving toward shared task definitions, rendering, and evaluation surfaces.
A closer look at this week’s progress 🧵
The Policy Checker Page is LIVE!
Remember we talked about showing our intermediate model and success heatmap? You can now check them live on our hub!
For each intermediate policy, you can:
- View its summary — success rate, task type, checkpoints, and more.
- Run it directly in your browser to observe its real-time inference. (If it looks a bit choppy, that's expected — the model is inferring live!)
- Explore its success heatmap — showing the initial states under which the policy successfully completes the task.
This gives the community a lens into our backend policy training, and lays the groundwork for better demonstrating our recover from failure training loop going forward.
How to enter ⬇️
Click-and-Drag Gripper Control is now LIVE
We've been simplifying high-hardware-demand, complex simulation teleoperation through web-based keyboard-and-mouse control — and now we're taking it one step further.
Direct click-and-drag gripper control is officially here.
This eliminates a significant number of intermediate steps. Simply reorient your viewing plane by adjusting the camera perspective, and focus only on gripper-object contact.
See it in action 👇
Physical AI onchain isn't just exciting—it’s accessible to everyone. 🦾
Huge thanks to @baseapac for the feature at SuperAI! We showed that simulation is no longer locked behind expensive rigs.
Zero special hardware needed. Anyone can train real robots directly on @base today.