Details new OpenAI agent primitives and how Computer Use got faster by combining screenshots with accessibility data, DOM, and generated code. Engineers building on the API can plan around async tools, steering, and higher-level agent state.
Key takeaways · AI-distilled
Dots gives every agent its own Linux computer in the cloud.
Weinstein says Computer Use can now complete some tasks faster than the average human, recovers from failures better than before, and gets richer context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → from App Shots than from ordinary screenshots.
Computer Use can close the loop between writing software and testing it, which makes it useful for coding, testing and QA.
Handa covers longer prompt cachingReusing the model's processed form of a repeated prompt prefix so subsequent calls skip re-reading it, cutting cost and latency substantially.Full definition → with cache pre-warming and server-side context compactionSummarizing an agent's earlier conversation to free room in the context window so a long session can keep going.Full definition → for long-running threads; OpenAI also uses the Decisions API internally for support classification.
Terms in this piece · Glossary
tool use — A model's ability to call external functions — run code, search the web, edit files — instead of only generating text.
context compaction — Summarizing an agent's earlier conversation to free room in the context window so a long session can keep going.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
prompt caching — Reusing the model's processed form of a repeated prompt prefix so subsequent calls skip re-reading it, cutting cost and latency substantially.
video🔬Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics)
video🔬 Google's AI Scientist Started as an Attempt to Automate Kaggle — John Platt, Google Fellow
videoA Worm With 302 Neurons Inspired Their Architecture — Ramin Hasani, Liquid AI
videoOrbs: Shifting Coding to Cloud — Quinn Slack, Amp Code
articleWhy Dwarkesh is Wrong about Computer Use + How OpenAI shipped its Jev competitor in 1 Weekwww.latent.space
articleImpressions from visiting OpenAI, Anthropic, & CursorGergely Orosz
podcast“Engineers are becoming sorcerers” | The future of software development with OpenAI’s Sherwin WuLenny Rachitsky