Free Lesson
Can AI agents improve themselves?
30 min
Oct 1, 2026 11:00 AM
Virtual (Zoom)
In this video
What you'll learn
Tell retries from persistent improvement
Use a simple test: what artifact changed, did it persist, and will the next run inherit it?
Inspect the improvement loop
Trace a prompt or configuration change through proposal, evaluation, selection, and persistence.
Find what the agent cannot improve
Identify the fixed evaluator, task inputs, and outer process that bound the reported improvement.
Why this topic matters
Claims of recursive self-improvement often group together retrying an answer, retaining experience, rewriting a prompt, and building a successor system. These are different mechanisms with different evidence. We will inspect the propose-evaluate-select-persist loop, map it onto a forecasting agent, and identify the evaluator and outer process that remain fixed.
