The Learner Brain™
AI That Learns Your Project
A methodology for structuring AI memory so your tools improve their own future behavior. Platform-agnostic — apply it across Codex, Claude Code, Antigravity, Cursor, Cline, Windsurf, and 13+ other AI coding assistants.
The goal is not better answers—it's fewer corrections over time.
"The Learner Brain is a meta-cognitive layer that turns repeated friction into permanent system upgrades."
The Dumb Agent Tax
Every time your AI says "Oh yeah, I forgot…" you just paid a tax. In tokens. In time. In focus.
Constant Retrying
Stateless systems can only retry. They can't evolve. You keep re-explaining the same things.
No Memory
If a system forgets what worked yesterday, it cannot compound. Memory is power.
No Compounding
Intelligence that doesn't persist is intelligence that doesn't compound into leverage.
The Four Components
Every Learner Brain is built from these four fundamental building blocks.
Rules
How I always behave. Persistent preferences and constraints that never need re-explaining.
View TemplatesSkills
How I do repeatable tasks. Multi-step procedures that become automatic behaviors.
View TemplatesWorkflows
How I approach complex work. Strategic sequences that orchestrate multiple skills.
View TemplatesMeta Layer
How I learn. The self-awareness that detects patterns and proposes upgrades.
Learn The PlaybookThe Three Triggers
When the Learner Brain proposes an upgrade
Repetition Trigger
"If it happens twice, it shouldn't be manual again."
Manual Labor Trigger
"If it's multi-step, it should become a skill."
Unwritten Law Trigger
"If the user corrects a preference, it becomes a rule."
Rules Propose. Hooks Enforce.
The meta-learning loop ends at "codify" — but a rule the agent keeps ignoring is still costing you corrections. Hooks close the gap: small scripts that fire inside the agent loop and guarantee the behavior.
Proposed
A rule in context. Flexible, cheap — and occasionally ignored.
Enforced
A hook in the loop. The command is denied, the check runs — deterministically.
Governed
Hooks are rules too — reviewed, versioned, and git-tracked like everything else.
A rule the agent keeps breaking graduates into a hook.
Explore the Enforcement LayerHow You Know It's Working
The signs of a healthy Learner Brain
You repeat yourself less
It asks to upgrade itself
You correct it less
Work speeds up over time
Behavior becomes predictable
Predictability is intelligence.
Works With Your Tools
The Learner Brain is a methodology, not a product. Apply it to any AI coding assistant that supports custom instructions. Pick the platform that fits your workflow — see the compatibility matrix on the Templates page.
Ready to Build Your Learner Brain?
Start with the Quickstart Guide—you can build a minimal viable Learner Brain in one sitting.
Stay in the Loop
Get updates on new patterns, templates, and AI development insights.