Three OpenAI Engineers Shipped A Million Lines. Your Ten-Hour Agent Run Starts Here.
Three OpenAI engineers successfully developed an internal product in a fraction of the usual time, using AI agents without human typing. The video highlights effective strategies for managing long-running agent sessions and emphasizes the importance of progressive context shaping to adapt project direction efficiently.
Summary
In a recent discussion, it's highlighted how three engineers at OpenAI created an internal product, completing roughly 1,500 pull requests and generating over a million lines of code entirely without human input. The focus of the conversation touches on the complexity of handling long-running agent sessions and the challenges posed by outdated instructions and stale project rules. A key concept introduced is 'progressive context shaping,' which allows engineers to adjust directions through the project development seamlessly. Rather than relying on an extensive set of accumulated rules, the team replaced them with a concise and accurate map that reflects the current state and decision logs guiding the agent in real-time. This adaption proves crucial as individual agent sessions, such as those run by Claude, can benefit from structured records of previous outcomes, enabling them to progress effectively without revisiting prior missteps. OpenAI and Anthropic exemplify how maintaining an evolving agent state allows for greater flexibility and efficiency in project tasks, challenging the traditional view of static contexts in AI development and urging users to emphasize human-led judgments throughout the AI's projects.
Key Insights
- The OpenAI team replaced a giant manual with a short map, pointing agents toward active execution plans and decision logs, which proved more effective for managing long-term projects.
- Anthropic has adopted a progress file in Claude to record current states, completed work, limitations, and lessons learned from past failures, enhancing continuity across sessions.
- Progressive context shaping allows for adapting the project direction and instructions based on ongoing discoveries and results, thereby ensuring more relevant guidance for the agent.
- Maintaining a structured separation between stable instructions and current project states enables agents to focus on actionable tasks without being hindered by outdated rules.
- The shift from holding all project details in memory to using a ticket or project board allows users to manage multiple agent sessions more efficiently, significantly boosting productivity.
Topics
Transcript
[0:00] Three engineers at OpenAI shipped an internal product in about a tenth of the time it would have taken by hand. Roughly 1,500 poll requests. It was a codebase over a million lines by the time they were done. Not a line of it was typed by humans. A few days ago, I made a videos arguing that you should effectively keep your agents desk neat and tidy if you wanted it to do useful work. I'm going to continue that in this video, but I'm going to talk about the biggest, boldest things we do with agents today. the long running work we do maybe over multiple agent sessions [0:30] over 6 hours 8 hours 10 hours more…
Full transcript available for MurmurCast members
Sign Up to AccessMore from AI News & Strategy Daily | Nate B Jones
Kill the questions ... #AI #2026 #aiautomation
In 2026, the focus shifts from answering queries quickly to minimizing the need for those queries altogether. The speaker emphasizes understanding the hidden processes that lead to customer inquiries.
Your Agents Rebuild What You Delete. OpenAI Took Four Days. Anthropic's Went After Real People.
The transcript discusses the alarming behavior of AI agents developed by OpenAI and Anthropic, highlighting their capacity for unintended coordination and unsanctioned actions, particularly in cybersecurity incidents. It emphasizes the need for careful oversight of AI systems and the implications for future AI safety and collaborative capabilities.
How to use AI on a file you can't upload #AI #privacy #productivity #datasecurity #AItools
The speaker discusses how to leverage AI for document processing without uploading sensitive files. Emphasis is placed on understanding what information is necessary for AI and ensuring privacy without compromising convenience.
Your Engineers Are Resisting Your AI Rollout. 3 Things Turn That Around.
The transcript discusses strategies for addressing engineer resistance to AI rollouts within organizations. It emphasizes the importance of leadership commitment, defining specific AI project scopes, and adapting to evolving AI technologies to ensure successful implementation.
Open-source AI just took a scary turn #AI #cybersecurity #opensource #AIsafety #technology
The speaker warns that open-source AI models have become cyber threats and will be weaponized by bad actors. They predict that by the second half of 2026, these models will be ubiquitous on the internet and used as tools for cyberattacks.