REWORK Bring your AI agents to Basecamp
72 snips
Mar 25, 2026 They explain making Basecamp accessible to AI agents via a Command Line Interface so agents can act faster than through browsers. They discuss treating agent-friendliness as accessibility and why speed and tight feedback loops matter. They describe agents turning plans into projects and how shared agent actions help teams collaborate. They outline cautious, reusable ways to expand agent access across other products.
AI Snips
Chapters
Transcript
Episode notes
Agent Accessibility Is Like Building A Ramp
- AI agent accessibility is like ramping a product for new users rather than redesigning how everyone uses it.
- David Heinemeier Hansson built a CLI so agents can interact via fast text loops instead of slow browser screenshots and image parsing.
Text Loops Unlock Agent Speed
- Text-based CLIs match LLM strengths and enable much faster feedback loops than browser automation with screenshots.
- Faster tokens-per-second lets agents iterate (try, adjust, retry) at near-human developer speed, making them practical tools.
Kimi K25 Built A Project In Seconds
- David tested fast open-weight models like Kimi K25 to design a Basecamp project and was amazed at how quickly it filled to-dos, messages, and schedules.
- The agent created assignments and schedule items, turning a text plan into an actionable project.