Sabitlenmiş Tweet
Dave Pitt
376 posts
Dave Pitt
@davehpitt
ML @Ultraroboticsco. Hustle like sam, grind like elon, body like dario
New York Katılım Haziran 2025
75 Takip Edilen92 Takipçiler
@cremieuxrecueil It’s the jammer world record. Low key unimpressive that the margin was this tight with the whole body tech suit.
English
Holy shit the record actually was 20.88 and this guy got 20.81.
The Enhanced Games actually did it!
Enhanced Games@enhanced_games
A $1,000,000 WORLD RECORD SWIM! Kristian Gkolomeev wins the Men’s 50m Freestyle in 20.81s and takes home $1,000,000 bonus + $250,000 first place prize and reclaims his 50M Freestyle world record.
English
People beat the jammer WR in full body suits all the time come on man is the goal to do something interesting or glaze big pharma
Trung Phan@TrungTPhan
Enhanced Games athlete in photo is James “The Missile” Magnussen. Australian swimmer with an Olympic silver medal. Said he’d “juice to the gills” if offered $1m to break 50m freestyle record of 20.88. Enhanced Games agreed. This vid of him is most viewed on its YouTube channel and top comment is incredible: “you can see his back from his front”. Absolute tank. 6’6 and will weigh ~250lbs for the swim. For the inaugural Enhanced Games this weekend, Magnussen will also wear a full-body polyurethane super swimming suit that was banned after 2008 Beijing Olympics (he’s been training and doping at a facility in LV).
English
@kyleanthony @lush_amorelli DO NOT trust an adult man who uses this emoji
English
Dave Pitt retweetledi
@aurel_arnold It’s possible your policy is just overfit to the data and the adherence from chunk to chunk is not as good as you’d want. Honestly i take it back, maybe plot some rollouts of your policy over an episode? Record teleop data, then run policy inference offline and plot the chunks
English
@davehpitt This should help with general smoothness of the policy in my understanding, but how about the jumps when handing over control back to the policy?
English
DAgger-style intervention now works with the Quest.
I sometimes get ugly “jumps” when handing over control to the policy though.
Caused by OOD states the ACT policy hasn’t learned smooth actions for yet?
Aurel Arnold@aurel_arnold
Now I know why it didn’t feel natural. Two issues: 1. DLS over all 6 joints couples position and orientation. Wrist motions lead to elbow swings. 2. WebXR reports controller pose somewhere at the palm, but you naturally rotate around your wrist. Fixed with Pieper-style decoupling + a 5-sec in-VR wrist-twist calibration.
English
Dave Pitt retweetledi
Come ship with us as we deploy robots faster than ever before possible.
Ultra is hiring across engineering, manufacturing, and operations:
— Senior Machine Learning Engineer
— Application Engineer
— Senior Mechatronics Engineer – R&D
— Senior Mechatronics Engineer – Robot Platform
— Supply Chain/Manufacturing Engineer
— Robot Assembly Technician
— Robot Lab Manager (2nd Shift)
— Robotics Field Support Engineer
— Robotics Software Engineer
Careers page linked in comments. ↓
English
Super exciting. Reliability and steerability are must-haves for an economically viable product
Physical Intelligence@physical_int
Our newest model, π0.7, has some interesting emergent capabilities: it can control a new robot to fold shirts for which we had no shirt folding data, figure out how to use an appliance with language-based coaching, and perform a wide range of dexterous tasks all in one model!
English
Dave Pitt retweetledi
Proud to see our collaboration with @physical_int featured on @LightconePod.
Earlier this year, we worked with PI to run their foundation model on our Gen 1 robots, deployed in real customer warehouses, autonomously working for full shifts. As we roll out Operator, we're excited to be pushing even harder on autonomy and performance.
Y Combinator@ycombinator
Physical Intelligence (@physical_int) is building a foundation model that can control any robot to do any task — what the team describes as the GPT moment for robotics. The company's cross-embodiment approach trains across many different robot platforms, and recent results show tasks being performed zero-shot that last year required hundreds of hours of data collection. In this episode of the @LightconePod , co-founder Quan Vuong (@QuanVng) sat down with @garrytan, @snowmaker, @sdianahu, and @harjtaggar to talk about why robotics is finally ready for its scaling moment, how PI runs its models in the cloud rather than on-device, and the playbook for what Quan sees as a Cambrian explosion of vertical robotics companies. 00:00 — Robotics just got cheaper 00:41 — The GPT moment for robotics 02:24 — Why robots didn’t work before 05:30 — The breakthrough that changed everything 09:12 — The data problem 13:33 — Robots learning without data 15:05 — Robots folding laundry (for real) 22:18 — From engineering problem → ops problem 29:12 — The startup playbook 38:46 — Thousands of robotics startups are coming
English
We built v0 of this robot last August. Really cool how far it's come :)
Ultra@Ultraroboticsco
Introducing Operator, our newest industrial AI robot built to work, not demo. Operator handles your warehouse's most repetitive tasks: packing, sorting, and kitting. Up to 24 hours a day, with flexibility and consistency that allows businesses to scale quickly. This is what we've been building. ↓ [1/4]
English
If you had two software engineering offers:
> One pays you $500k/year salary, but covers zero LLM tokens.
> One pays you $400k/year salary, but gives you $500/day free LLM tokens.
Which one are you taking?
English