Self-improvement loops only pay off where the target function is precise; coding set a misleading precedent because compilation hands you a free signal. This shows the task specificity required and where domain experts must stay in the loop.

• videoBuilding Great Agent Skills: The Missing Manual
• videoSkill Issue: How We Used AI to Make Agents Actually Good at Supabase — Pedro Rodrigues, Supabase
• videoSkills are the New SDKs - Elvin Aghammadzada, DataRobot
• videoHow I deleted 95% of my agent skills and got better results — Nick Nisi, WorkOS

• videoLoop Engineering from First Principles — Kyle Mistele, HumanLayerAI Engineer
• videoImproving Agents is a Data Mining Problem — Vivek Trivedy, LangChainAI Engineer
• videoAgents, codebases, and teams — Aditya Khandelwal, Amazon AGI LabAI Engineer

Checking sign-in…

Loading comments…