---
title: "Chapter 15: The No-Weights Complement"
description: "Reflexion-style notes and skill accretion — learning mechanisms that run in parallel with training and cost zero GPU."
type: lesson
order: 15
chapter: "05-the-flywheel"
---

# Chapter 15: The No-Weights Complement

This runs in parallel with everything else in Part 5 and costs zero GPU.

## Reflexion-Style Notes

After each run the agent writes a lesson to a persistent file, and reads that file at the start of the next run. Instant learning, no training.

## Skill Accretion

When the agent solves something reusable, it writes a script or a `SKILL.md` and keeps it. The library grows. Experience becomes tooling.

## Why This Belongs in a Fine-Tuning Course

These stack on top of the tuned model. The fine-tune makes the agent competent at *the loop*; the notes make it competent at *your repo, this week*. Different timescales, different mechanisms, no conflict.

This is also the resolution of the course's third breaking-point warning: codebase facts belong in context, behavior belongs in weights. Notes and skills are the context side of that split, done deliberately instead of accidentally.

---

[← Chapter 14](/fine-tune-local-agent/05-the-flywheel/14-learning-from-failures/) · [Next: Chapter 16 →](/fine-tune-local-agent/05-the-flywheel/16-scale-up-and-ship/)