Exciting cause and stellar team.
💥 Introducing Warp Sciences, a new company applying AI and robotics to help the world develop new medicines faster.
We’ve raised a $215M Series A led by @lightspeedvp to bring frontier AI out of the browser and into the lab, and build scientific capacity here in America.
I co-founded Warp with my long-time friend @dpatil, FAIR co-founder @manohar_paluri, and drug hunter Sol Reisberg. The early team is full of incredible people like @bkspears9, who helped achieve Q>1 fusion at LLNL.
Peter Welinder retweeted
We just launched Isaac Lab 3.0! I wanted to take a minute to share what this one means to me, because the story goes back a while.
Back in 2015, during my PhD, I happened to stumble across the DQN paper in Nature. I was familiar with the ML space at the time, but it was the first time I really had any visibility on RL. I came from a hardcore electro-mechanical and controls background, and I remember reading it and thinking, this is some of the coolest work I’d ever seen. It completely shifted what my research goals looked like. I still did a lot of hardware design, and I led a team to design some crazy shape-shifting UAVs, but that’s around the time I started obsessing over simulation and RL.
That change in headspace led me to join @OpenAI's robotics team in 2019 and the Rubik’s Cube project, which was all about scaling sim-to-real with RL on dexterous manipulation tasks. We had a huge CPU cluster running MuJoCo-C for rollouts, plus a very large V100 cluster for async rendering. It was a really sophisticated setup, and it was the first time I witnessed the power of scaling RL.
During this time, I vividly remember a backyard barbecue at @woj_zaremba’s house with @npew. He showed me his phone and said, “Hey, our friends at @nvidia have a new simulator, and they said they can train entire Rubik’s Cube policies on a single GPU.” I didn’t know it at the time, but what I was looking at was an early build of Isaac Gym.
Little did I know that years later, through a very serendipitous set of events, I’d end up working with some of the same people who built it. @gavrielstate, @kelly_guo, @milesmacklin and so many of the people behind that OG simulator everyone loved are the team behind Isaac Lab 3.0. And for the last year, I’ve had the privilege of leading product for Isaac Lab alongside them.
Lab 3.0 is a huge, huge achievement for the team and for robotics at @NVIDIARobotics. It unlocks a whole bunch of new capabilities, it’s lightweight, and we really built it to be something people love. We’re already getting strong support from customers, and it’s gaining a bunch of positive traction. I’m incredibly proud of it.
I truly believe simulation can unlock incredible robot capabilities, and I’m so excited to keep building simulation products people love. I hope you all enjoy this release. There’s a lot more coming, and we’re really thankful for your support.
We’re excited to announce the official release of NVIDIA Isaac Lab 3.0. 🎉
Isaac Lab 3.0 brings a new modular and lightweight architecture along with major new capabilities across the robot learning workflow: multi-physics and multi-rendering support, training at scale, and streamlined deployment on real hardware.
With Newton integration, the new OVRTX renderer, RL post-training, and LEAPP deployment, there’s a lot to explore.
Here’s a look at what’s new 🧵
(1/7)
Peter Welinder retweeted
Sol completes the tasks less than 5 percentage points off Opus, but at less than a tenth of the cost!! $0.80 vs $10.53. Incredible performance from @openai 💪
Agents on Rails is back with Stage 3!
We put the best-performing models to the test with 7 new multi-step tasks, each spanning 3–8 tickets. From building Rails apps from scratch to navigating existing (and at times messy) legacy codebases, we wanted to see how well agents handle the kind of work developers actually face: going back and building on their own work.
Some highlights:
- Best overall: @OpenAI GPT-6.1 Sol offers a strong balance of accuracy, speed, and cost (60.8% success, $0.80 & 23m 35s per run)
- Most accurate: @AnthropicAI Opus 5.5 at max effort (65.3%, but it is also the most expensive and amongst the slowest)
- Fastest: @OpenAI GPT-6 Astra at medium effort (16m 36s, 54% success). Astra's max effort is also fast (22m 36s, 62.3% success).
- Cheapest: @OpenAI GPT-6 Luna ($0.19/run, 43.5% success)
We also combined all tasks from all stages into one central leaderboard.
Read the full Stage 3 report from @evilmartians here: rubyonrails.org/2026/10/13/a…
ALT Agents on Rails scoreboard showing stats for all 11 models tested in Stage 3 in October 2026. Claude Opus 5.5 (xhigh) is in first place with a 65.3% success rate.
Regulated scarcity.
These NYC apartments are standing empty, not because the owners don't want to rent them, but because rent stabilization laws prohibit them from charging enough to earn back the costs of necessary repairs.
My latest for @WSJFreeEx
Peter Welinder retweeted
🌶️ fixed it
only a fool would say AI is *creating* vulnerabilities faster than ever... the vulnerabilities were there already
this is a chart about us *fixing* vulnerabilities faster than ever before... it's a chart about the internet becoming secure.
Peter Welinder retweeted
OpenAI’s math release is, in the words of Opus, “the single most consequential mathematical release ever” *
LLMs have how made substantial progress on 4 of 7 Millenium Prize problems: Navier-Stokes (claimed), Riemann, Hodge and Birch-Swinnerton-Dyer. Poincaré was solved in 2003. The two left are P v NP and Yang Mills.
*conditional on verification
On average, each OpenAI result used only 3hrs of thinking compute on their new unreleased models.
Two years ago, models said 9.11 > 9.9. Today, we have results that the smartest human minds have not been able to achieve in their entire lives. It is clear that data, compute and algorithms scale. Every model generation (3mos) has made substantial intelligence progress.
It also becomes incredibly hard to not believe that all knowledge work will change monumentally over time. The things AI can not do in the foreseeable future are very likely context-bound (don’t have access to the right information) than intelligence bound. Some might argue they are also creativity-bound or judgement-bound (what should I work on), although it can be argued that future generations could solve for this (given, say, the advancement in research taste for models over time).
AGI is defined as surpassing human capabilities on virtually all cognitive tasks. By most interpretations of that definition, we are there. The domains humans are still better than AI, such as, robotics / physical world control (data-bound?), natural science research (data-bound), some creative domains like writing, movies, music (creativity-bound), super long tasks (context-bound), choosing problems to solve (creativity-bound) and maybe human relationships management (meat-proxy bound?).
In many ways, we have achieved AGI.
Peter Welinder retweeted
Muse is extremely user-friendly, Claude is very smart and has great design taste, but in my experience for research tasks ChatGPT smokes them both.
Peter Welinder retweeted
I just called Chungus (my OpenAI Dot) on the phone while picking up mexican food and asked it to transfer some Rex benchmarks from last night on my Mac Studio to my MacBook Pro via Tailscale and I just got to my desk and there they are. What a fucking time to be alive.
Peter Welinder retweeted
Gesture recognition using the Decisions API by @OpenAIDevs.
Thanks to it's vision understanding it is way more accurate than my previous Jev-based demo. It was also faster. cc @thsottiaux @jxnlco @pvncher @tldraw @steveruizok
Peter Welinder retweeted
We can reindustrialize America around high-scale manufacturing, superintelligence, and robotics.
We can be a nation made solvent through taxing exports, not taxing its citizens.
We can eradicate most diseases, energy shortages, hunger, housing insecurity, and violent crime.
Best growth and prosperity strategy for countries: cheap energy, high-skill immigration, and don’t over-regulate. Pretty simple.
So much worry about China sprinting ahead on foundation models. I would assume China doesn't want their own models hacking into their systems either.
In other words, once you hit the frontier, you have no choice but to be careful and focus on safety.
Peter Welinder retweeted
Just this morning, needles and other drug paraphernalia reported next to Marshall Elementary.
"A couple weeks ago, a first grader reached through the fence surrounding the recess yard and grabbed a bloody hypothermic needle. He passed it around to four other children before it was caught."
Parents and staff of Marshall Elementary in the Mission District and the area’s supervisor said street conditions have worsened since school started. They’re asking for more help. sfchronicle.com/sf/article/s…
