Corey Lynch (@coreylynch) / X

Corey Lynch

577 posts

Corey Lynch

@coreylynch

Director of AI at @figure_robot, building Helix 🧬

San Francisco, CA

Joined January 2009

Pinned
Corey Lynch
@coreylynch
May 20
We just crossed a full week of non-stop autonomous work with Helix-02. 24/7 operation, >200k packages handled with a pixels-to-actions neural network.
Figure
@Figure_robot
May 20
Entering Day 8 of the livestream: 167 consecutive hours and 209,000 packages sorted autonomously x.com/i/broadcasts/1…
7.1K
Corey Lynch
@coreylynch
Mar 13, 2024
We are now having full conversations with Figure 01, thanks to our partnership with OpenAI. Our robot can: - describe its visual experience - plan future actions - reflect on its memory - explain its reasoning verbally Technical deep-dive 🧵:
Figure
00:00
Figure
670K
Corey Lynch
@coreylynch
Feb 20, 2025
My team has been working for a year trying to solve intelligence for the home Meet Helix 🧬: the first Humanoid Vision-Language-Action model This is one Helix neural network running on 2 Figure robots at once They've never seen any of these items before
Figure
00:00
Figure
275K
Corey Lynch
@coreylynch
Aug 15, 2025
Zero new data or retraining of any sort btw. Just ask helix “fold the towels”.
00:00
Noam Brown
@polynoamial
Aug 12, 2025
Replying to @kimmonismus
Does it still work if you raise the table 6 inches?
140K
Corey Lynch
@coreylynch
Sep 3, 2025
“Helix, load the dishwasher”. No algorithmic changes, just new data.
00:00
63K
Corey Lynch
@coreylynch
Oct 7, 2025
We have something exciting to share soon
Figure
00:00
Figure
28K
Corey Lynch
@coreylynch
Sep 18, 2025
Announcing Project Go-Big There’s no YouTube for robot data, so we’re building one ourselves - Figure now has access to >100k real homes via Brookfield - We’ve begun scaling human collect - We already have initial results: direct human video-to-robot transfer to Helix 🧬
00:00
149K
Corey Lynch
@coreylynch
Aug 20, 2025
Our fully sim2real RL whole body control is getting quite stable
Brett Adcock
00:00
Brett Adcock
52K
Corey Lynch
@coreylynch
Mar 13, 2024
Replying to @coreylynch
Finally, let's talk about the learned low-level bimanual manipulation. All behaviors are driven by neural network visuomotor transformer policies, mapping pixels directly to actions. These networks take in onboard images at 10hz, and generate 24-DOF actions (wrist poses and
00:00
68K
Corey Lynch
@coreylynch
Mar 13, 2024
Replying to @coreylynch
Let's break down what we see in the video: All behaviors are learned (not teleoperated) and run at normal speed (1.0x). We feed images from the robot's cameras and transcribed text from speech captured by onboard microphones to a large multimodal model trained by OpenAI that
79K
Corey Lynch
@coreylynch
Jan 7, 2024
Incredibly excited to share some recent progress :) What you see in the video: ⁃A learned, end-to-end visuomotor policy mapping onboard images to low level actions at 200hz. ⁃All behaviors (including corrective) are fully autonomous (not teleoperated). ⁃1x speed. I’m
Brett Adcock
@adcock_brett
Jan 7, 2024
Figure-01 has learned to make coffee ☕️ Our AI learned this after watching humans make coffee This is end-to-end AI: our neural networks are taking video in, trajectories out Join us to train our robot fleet: figure.ai/careers
Figure
00:00
Figure
88K
Corey Lynch
@coreylynch
Aug 12, 2025
Here's Helix folding laundry. I'm consistently surprised at how little we need to do to scale to hard new tasks. Zero code or hyperparameter changes, just new data.
Figure
00:00
Figure
19K
Corey Lynch
@coreylynch
Apr 25, 2017
Result of a very fun @GoogleBrain residency: Self-supervised robotic imitation of human motion arxiv.org/abs/1704.06888
GIF
Corey Lynch
@coreylynch
Mar 13, 2024
Replying to @coreylynch
Even just a few years ago, I would have thought having a full conversation with a humanoid robot while it plans and carries out its own fully learned behaviors would be something we would have to wait decades to see. Obviously, a lot has changed :). In my mind, Figure is the
26K