Agent Frameworks Considered Harmful — Rémi Louf, .txt

By ai.engineer

Categories: AI, Tools

Summary

Agent frameworks are a trap: the CEO of .txt spent two weeks building agents and discovered existing frameworks force you to edit prompts in code rather than manage them declaratively. YAML-based, event-driven agent systems with scheduled triggers prove far more practical than current framework abstractions.

Key Takeaways

  1. Agent frameworks waste developer time by embedding configuration in code. Switch to declarative YAML for agent definitions—enables version control, PR reviews, and instant deployment without rebuilding.
  2. Event-driven agents outperform cron jobs for automation. Trigger agents on system events (e.g., voice note uploaded) rather than time schedules to create truly autonomous workflows that react to state changes.
  3. Current agent UIs are transitional friction. Text-based interfaces and remote-controlled agents (via phone apps) are absurd stop-gaps—true agents should run unattended in background like autonomous robot mowers, requiring zero human intervention.
  4. Build the dumbest thing that works first. Louf rejected established agent frameworks to experiment from first principles, discovering simpler primitives (YAML configs + event scheduling) solve real problems better than heavyweight abstractions.
  5. CEO-level AI research is justified for early-stage founders. Taking 2 weeks to deeply understand emerging tech (triggered by Claude Opus capability jump in December) shaped product strategy and identified what existing tools actually lack.

Related topics

Transcript Excerpt

[music] Hi everyone. So originally I thought I was going to give a very technical talk but I saw was in the leadership track which I'm not sure what it means but I was like okay I'm going to do half high level and half technical. So it's more a story about what I you know what I did in January because in around December um agents kind of became really good uh you know there was a step function something happened uh I think it was opus 4.6 six and that's when I realized and I work in AI where I was like okay this thing is really happening and so I took two weeks out I took two weeks away so I'm the CEO of text uh you know which is 15 people company I just told my CTO I was like okay I'm just going to go away for two weeks and I'm just going to dive in this thing and try to understand what w…

More from ai.engineer