Peter Welinder

Devices @OpenAI

[Peter Welinder @npew]

Todo lists for agents. Feels like the right way to do things in 2026.

[Jan Senderek @jans]

In 2014, I sold my first company Loom to Dropbox.

Today, I’m proud to announce Roma 2.0 (YC F26): the first to-do list for AI agents.

Everyone uses a different agent: Claude, Grok Bot, Muse…

Until now, none of them could work together.

Roma automatically creates tasks and hands them to the right agent.

Allowing you and your agents to work together, with shared context.

Try it here: https://t.co/w9Ag9OrTUO

[Hayden Field @haydenfield]

In many circles, “Ask Chat” has replaced “Google it,” and that’s reflected in a new announcement from OpenAI: the volume of search queries run through ChatGPT now equals 20% of the volume run through Google. https://t.co/Jfaal7r8JI

[Peter Welinder @npew]

@oegerikus and his team were early visionaries who saw the potential of LLMs for coding early. Building Github Copilot together and solving all the little things we take for granted today -- from optimizing latency and caching to coding evals -- was so much fun.

[Oege de Moor @oegerikus]

The Origin Story of Github Copilot:

In 2020, @GitHub had just acquired my startup, Semmle, and the CEO, @natfriedman, gave me free rein to find our next product.

At the time, I had a bee in my bonnet: GitHub had more data than anyone else about developer behavior, but it wasn’t using that data to build new products.

I started exploring how to change that.

In June 2020, I got early access to GPT-3 and saw its potential to write code, especially if trained on the code on GitHub. It was the obvious project to work on.

It could generate a little code on its own, but almost always got it wrong. So I gave the model a set of 50 programming puzzles, the kind I used to set at Oxford entrance interviews.

I asked the model to solve each puzzle 1,000 times, using automated tests to check it. Given enough attempts, it got almost all of them right. That's when I knew it would be able to code.

I assembled a small initial team of ~10 people to work on what became GitHub Copilot. The first two people who joined were co-founders of the project: @thewunderalbert and @alexgraveley

Fortunately key people at OpenAI shared our vision, and it was the perfect partnership at the perfect time with @sama, @gdb, @ilyasut, @DarioAmodei and Sam McCandlish (months before Anthropic was born!), @npew and several others.

OpenAI trained the model behind it, Codex, and we built the product around it. It ran inside VS Code and suggested lines of code and functions as you typed.

The VS Code team, led by @ErichGamma, moved mountains to give us exactly what we needed. The Azure AI team of Misha Bilenko also helped make it scalable.

When we launched the technical preview in June 2021, it hit 1.2 million users in 12 months. It was an instant hit with developers and helped write billions of lines of code.

It also left me with an uncomfortable thought. The models had learned to code from public source code, and public source code has a lot of vulnerabilities in it.

It was easier than ever to write vulnerable code, and developers were shipping it faster than ever.

But, I knew that if AI could write the code, it could find the holes in it too. It was imperative that defenders got it before attackers did.

In 2023, I left GitHub to build @Xbow to do exactly that. Since then, we've already helped our customers find and fix more than 14,000 zero-days and the work has only just begun.

The same team that started the AI coding revolution is now leading the charge on AI security.

[DEJAN @dejanseo]

Anthropic: We have the best model.
Google:

[Peter Welinder @npew]

Incredibly impressive.

[Alexander Koch @alexkoch_ai]

You can now book our humanoid cleaners on a recurring subscription, for up to 3 hours per cleaning at https://t.co/Brl42KyISx.

Video is first take, uncut, shown at 4× speed. See the reply for detailed footage at 1× speed.

[Peter Welinder @npew]

@bengoodger Intrigued! Will check it out.

[Peter Welinder @npew]

Important to give the cars a good view of the water.

[Mark R. Brown, AICP, CNU @Car_Free_USA]

This is what happens when we give traffic engineers too much power.

[Hamido Teck @HamidoFx1]

GPT 6.1 Terra is absurdly good at 3D modeling and motion.

I asked it to build an interactive showcase that counts to 5 and explodes a mechanical hand down to every single gear, wire, and panel.

We are living in the build era.

[Andrey Kurenkov @andrey_kurenkov]

bencmaxxers getting absolutely wrecked , you love to see it

[Jeff Harris @jeffintime]

This began with GPT-5. We kept hearing: “love the model, so slow.” And it wasn’t TBT. The model just made too many tokens.

This is what happens after spending a full model generation obsessing about token efficiency.

[adam @adamdotnew]

GPT 6.1 Terra is blazing fast at CAD!

It is now our default model for free users.

Watch it work in real time.

[Peter Welinder @npew]

We’re all living in a simulation.

[Sundar Pichai @sundarpichai]

Excited to share Genie 4 today our most realistic and interactive world model:

- steer dynamic environments in real time, shaping events, camera angles, and actions

- generate native audio matched to the world, from realistic background sounds to characters who can converse in real time!

- simulate photorealistic scenes and real-world physics

- create more coherent multi-step events through reasoning from Gemini

We've started to roll out as a research preview to trusted testers. Coming soon to Google AI Ultra subscribers in Project Genie and early access partners via the Gemini API and AI Studio.

[tobi lutke @tobi]

This video / song is criminally overlooked 😂

[Greg Brockman @gdb]

trying to rise to the occasion. grateful to be working with a team that feels the responsibility as deeply as i do.

[Peter Welinder @npew]

Exciting cause and stellar team.

[Kevin Weil 🇺🇸 @kevinweil]

💥 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.

[Ruben D'Sa ✈️ CoRL 2026 @ruben_dsa]

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.

[Freddy Vega @freddier]

GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right.

A team checked. And yes. GPT-6 Astra was right. New score: 98%

[Peter Welinder @npew]

@manuaero Congrats!

[DHH @dhh]

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 💪

[Peter Welinder @npew]

Regulated scarcity.

[Emma Camp @emmma_camp_]

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