"How would you build a biodefense tool?" I thought @velvetatom was asking hypothetically ...she was not. So excited to see Valthos finally out of stealth – the work is fascinating. I trust Tess & team deeply to push biodefense forward with the seriousness and technical rigor it

👀 Life update: Today I’m announcing Rehearsals! @RunRehearsals watches every user and tells you why people pay, why they leave and what exactly to build next. So why now? • Big companies are getting smaller and small teams are doing more. • Multimodal models (especially

passes the vibe check

Today we’re introducing GDPval, a new evaluation that measures AI on real-world, economically valuable tasks. Evals ground progress in evidence instead of speculation and help track how AI improves at the kind of work that matters most. https://t.co/uKPPDldVNS

agents are reshaping how we work, reason & operate. so, we're hosting a design eng fireside to discuss the future of agentic interfaces ft. @ilanbigio (openAI), @annarmonaco (paradigm), @liamfmatteson (browserbase), @silasalberti (cognition) & @JasonBud (cursor)

Paradigm is the AI-native spreadsheet to eliminate menial work. Thousands of users have saved 10,000+ hours with Paradigm, and you can be next. Get your first month free today, then plans start at just $20/month.

- https://t.co/A5wudwZVG8 - https://t.co/9XbGkwBoks - cookbooks: https://t.co/FjjMom4RYo (harmony, raw cot, fine-tuning, ollama, etc) HUGE kudos to the team, and especially @DKundel for creating so many of these himself. i will be buying you a beer

along with the gpt-oss launch 🎉 we shared some quality resources i'm surprised aren't getting more attention: - playground: gpt-oss [dot] com (!) - harmony: how we tokenize chat, CoT, tool calls - raw CoT - fine-tuning - ollama ...and so many cookbooks. links to all below 👇

So excited to be launching 4 features in the OpenAI API today: 1. Deep research with support for web_search and remote mcp (for your data that's not publicly accessible on the internet) 2. Web search in o3, o4-mini, and o3-pro — at a new reduced price of just $10/1K queries 3.

https://t.co/khXmq2Rgoh

a great way to understand each method is to ask: what is the model learning? SFT - learns a soft mapping DPO - learns a nuanced "delta" or direction in high-dimensional space RFT - learns to reason, figuring out the thought process necessary to accomplish a task

you usually don't need fine-tuning. but when you're pushing the boundaries of performance and efficiency, you do 🧠 RFT - train reasoning models on SotA tasks ⚖️ DPO - a/b optimization, tone matching ⚡️ SFT - distillation, formatting, efficiency when fine-tune w/ @openai? 👇

our next @openai build hour will cover a new RealtimeAgent feature for voice agents & improvements to our speech-to-speech model. as always, we'll be live building. tune in tuesday at 10am PT!

https://t.co/FObjotaOxV

thanks @stainlessapi for hosting the mcp panel last week! (and not booting me off for some warm takes) overheard: "down to earth", "spicy", "where's the bar" check out the full 1hr panel in the link below @RattrayAlex @mc_anthropic @shashankgoyal95 @TomHacohen @danshipper

How to set up a remote MCP server and connect it to ChatGPT deep research:

We just released the OpenAI Agents SDK for TypeScript and with it comes native support for Realtime Agents both in the browser and on the server! Come and join my @aiDotEngineer workshop later today at 3:30pm to get your hands on the new SDK https://t.co/gQLIsAyopZ

Nice slide from @ilanbigio from @OpenAI at @aiDotEngineer conference - Supervised Fine-tuning Give the model input-output examples and it learns to copy the pattern. Great for straightforward tasks like formatting or classification. - Direct Preference Optimization Show the

🚀 It’s live! Our free 2-hour course AI engineering: Agents is out now. Build production-ready AI agents with tool-chains, memory, multi-agent orchestration & evals. Built with @OpenAIdevs, taught by @ilanbigio + @ash0ts. Start learning now ⤵️

(huge fan of your work btw - just poking around :D)