AI Tools Review

Anthropic Ran 1,000 AI Agents For 8 Days. Yours Still Needs The Same Thing Explained Again.

Nate B JonesNovember 20, 202610:00

How do you get long-running AI agents to do the work you actually want, without explaining the same thing over and over? OpenAI and Anthropic report AI agents working for hundreds of hours and for days. I break down what keeps those long runs on track, for you and for your team. Full post: https://natesnewsletter.substack.com/p/long-running-ai-agents?utm_source=youtube&utm_medium=video&utm_campaign=long-running-ai-agents-launch&utm_content=description Nate's Library MCP (every prompt kit and guide, inside the AI you already use): https://unlock-ai.natebjones.com/guides/how-to-connect-nates-library?utm_source=youtube&utm_medium=video&utm_campaign=long-running-ai-agents-launch&utm_content=description My Links πŸ”— πŸ‘‰πŸ» Newsletter: https://natesnewsletter.substack.com/?utm_source=youtube&utm_medium=video&utm_campaign=long-running-ai-agents-launch&utm_content=description πŸ‘‰πŸ» X: https://x.com/natebjones πŸ‘‰πŸ» TikTok: https://www.tiktok.com/@nate.b.jones πŸ‘‰πŸ» Instagram: https://www.instagram.com/nate.b.jones What's really happening when an AI agent works all night? The common story is that longer runs just need a smarter model, but the real question is whether your agent has the current decisions, examples, and instructions behind the work. In this video, I share the inside scoop on long-running AI agents: - Why my research run produced a thousand questions I didn't need - How treating project information like code keeps agents on current decisions - What shared skills and permissions change when many agents touch the same work - Where to start alone with one job you keep explaining An overnight run can save you a day or cost you one, and the difference is the structure you give it. Chapters: 00:00 OpenAI ran AI sessions for hundreds of hours 00:52 An all-night run can leave a full day of cleanup 01:25 My research run that went sideways 02:04 The assignment was the problem, not the data 03:27 Getting the right context in front of the AI 05:14 Treat project information like code 07:52 Skills built for the team, not for you 10:57 Roles and permissions in a shared repo 12:53 Avoiding collisions when agents share work 14:39 How to judge whether it is working 17:28 Start alone with one job you keep explaining 19:20 Why team leaders need to see this Listen to this video as a podcast. Spotify: https://open.spotify.com/show/0gkFdjd1wptEKJKLu9LbZ4 Apple Podcasts: https://podcasts.apple.com/us/podcast/ai-news-strategy-daily-with-nate-b-jones/id1877109372

Watch on YouTube

πŸ€–AI Summary

Nate B Jones explains what keeps long-running AI agents on track, drawing on OpenAI and Anthropic reports of multi-day runs: current decisions, examples and instructions matter more than a smarter model.