


Head of Travel @ OpenAI. Founder of Odessia (acquired by OpenAI), founder of Sonder (took public).


Massive leap along the pareto frontier
Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. openai.com/index/introducing…

We’re hiring elite software engineers to join OpenAI to build the future of travel on ChatGPT, based in SF and working with @radoshi and I directly. Must be passionate about travel and a 3 sigma outlier on quality, speed and teamwork. openai.com/careers/full-stac…

This week my company was acquired by OpenAI and I now lead travel on ChatGPT. What would you like to see on the app? Use cases you'd love ChatGPT to nail when it comes to planning a trip?

And that’s before accounting for benchmaxxing
There's a belief that open source tokens are "cheaper" yet closed models dominate the pareto frontier at any level of intelligence. Price-performance improves so fast that free weights aren't enough to compensate for being as little as a few months behind.

Wasn’t even in my consideration set
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…

This has become a critical part of my workflow
openai dot and a 24/7 mac mini have made my life 15% better. couldn’t recommend more. will write a blog post on my setup

A related question: even if you can post train an open source model to outperform at a given task, how long before a general purpose pareto-efficient model comes out that does it better/cheaper? Paybacks better be short!

There's a belief that open source tokens are "cheaper" yet closed models dominate the pareto frontier at any level of intelligence. Price-performance improves so fast that free weights aren't enough to compensate for being as little as a few months behind.

Excited for higher quality social sciences
By every measure, Opus 4.8 beats human experts as peer reviewers in a study of 124 Harvard social science papers covering a range from statistical analyses to formal models to qualitative work. As academics, we need to figure out how to adapt the practice of science to AI, fast.

I'm excited to launch Relational Capital, a $50m solo GP fund. I spent the last 8 years as a founder, building @usekojo into the largest US construction procurement platform. Now I'm backing founders at the earliest stages with $1m-$5m checks. An abundant future lies ahead 🇺🇸

Maria's an insanely talented founder and now back to investing – couldn't be more proud of my exceptional wife!!
I'm excited to launch Relational Capital, a $50m solo GP fund. I spent the last 8 years as a founder, building @usekojo into the largest US construction procurement platform. Now I'm backing founders at the earliest stages with $1m-$5m checks. An abundant future lies ahead 🇺🇸



We're heading toward composing entire canvasses just in time, perfectly suited for the task. And even better when this is activated via a continuous voice interaction with slides animating in and out
GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone. Intelligent UI in ChatGPT delivers fast, interactive answers that make everyday questions more visual, complex topics easier to grasp, and interactive tools for your task available on the spot.

We're giving agents the same tools that professionals have. Their own phone number, their own email. The level of 24/7 personalized service and support consumers are about to get is game changing