Recursive self-improvement · RSI v6.2
Understand the AI that helps improve AI.
From concepts to your first supervised cycle. Compare evidence, recognize limits, and test improvements in support, education, and documents.
start now →18 lessons · 6 modules · about 232 minutes · free · phone and computer
What you start doing
Who it’s for
It’s for you if
- You work in management or education and want to evaluate and improve routines with AI.
- You have little time and no technical background.
- You prefer learning by doing.
It’s not for you if
- You already master the subject and are looking for something technical.
- You want a traditional video lesson.
Six modules, one practice per lesson
Module 1 · Basics: recognize what changed
Distinguish RSI, response reviews, and claims without evidence.3 lessons · 39 min
Module 2 · How it works: draw a cycle
Goals, roles, evaluation, memory, and adopting improvements.3 lessons · 39 min
Module 3 · Research: read real cases
AlphaEvolve, DGM, and the pieces of automated research.3 lessons · 39 min
Module 4 · Assessment: check before trusting
Held-out cases, manipulation of measurements and generated data.3 lessons · 39 min
Module 5 · Applications: use at work
Customer service, education, and document summaries with references.3 lessons · 37 min
Module 6 · Final project: decide with evidence
Choose an implementation, plan a pilot, and report its limits.3 lessons · 39 min
Research with source and context
AlphaEvolve, Darwin Gödel Machine, AI Scientist, DSec and evaluation studies. Original sources included with each lesson, consulted on 25/09/2026.
Marina and Ícaro are fictional characters. The screens illustrate comparisons; they do not show measured results from tools.
The practices teach supervised improvement. They don’t promise to create an autonomous AI or to train a new model.
Read the research and editorial decisions (Portuguese) · Download the experiment worksheet · Download practice cases · Download the approved knowledge worksheet