The learning method · Deeper insight
The Agentsable learning method.
This is a cycle, not a syllabus. It exists independently of any course we happen to be running, and every Agentsable program is built on it. Each stage below opens for examples and reasoning.
- 01
Understand the theory
Short, unglamorous explanations of how the system actually works, enough to reason about it, not enough to pass an exam.
Why it's in the cycle
We teach mechanism before interface: what a context window costs you, why retrieval changes the failure profile, what an agent loop is actually doing when it 'decides'. Roughly fifteen minutes of theory per hour of work, never more.
- 02
Watch real, unedited work
You see someone build the thing in full, including the parts that go wrong and the time spent stuck.
Why it's in the cycle
Edited tutorials remove exactly the material you need: the dead ends, the misread error, the moment of choosing between two reasonable designs. We leave those in, because that is the content.
- 03
Apply it immediately
Within the same session you build the equivalent thing yourself, from a blank file, on your own problem where possible.
Why it's in the cycle
The gap between watching and doing is where the illusion of competence lives. Closing it inside the same sitting is the single highest leverage design decision in the method.
- 04
Inspect the failure
Your build breaks. That is the assignment. You diagnose why, in writing, before you fix it.
Why it's in the cycle
We ask for the diagnosis first because the fix is worthless without it. A learner who can name the cause can transfer the lesson; a learner who only has a working fix cannot.
- 05
Get specific feedback
A reviewer responds to the decision you made and defended, not to whether the output looked impressive.
Why it's in the cycle
Feedback is on judgment: was the human checkpoint in the right place, was the failure mode identified, was the check cheap enough to actually run. Polish is not graded.
- 06
Improve and repeat
You revise the same artifact with what you learned, then take the next problem, which is harder.
Why it's in the cycle
The cycle is the product. A program is just a particular sequence of problems fed through it. When our programs change, this does not.
What we can and can't show you yet
We would rather show artifacts than describe them. Real learner builds, review threads, and failure write ups will be published here as participants agree to share them. Until they exist publicly, treat the descriptions above as our stated standard rather than as evidence.
Who will you practise this alongside?