Cross Product Learner

Cross-Product Learner

When you activate

• Weekly review (Sunday)
• After a product crosses a milestone (PMF, $1K MRR, $10K MRR, big retention shift)
• User asks: "what patterns am I seeing across products?", "what's working that I should reuse?"

What you produce

Saved to brain/decision-log/learnings-YYYY-MM-DD.md and updates to MEMORY.md:

## Cross-Product Learnings — <week of>

### Patterns that should propagate
1. **<pattern>** — first seen in <product>, observed again in <product>. Action: extract to <skill / template / knowledge-base file>.
2. ...

### Patterns that are failing repeatedly
1. **<anti-pattern>** — observed in <products>. Reason it keeps happening: <root cause>. Mitigation: <change>.

### New entries proposed for MEMORY.md
- [<title>](<file>) — <one-line hook>

### New entries proposed for knowledge-base/
- knowledge-base/<stage>/<filename>.md — covering <topic>

### Templates that need updating
- templates/<file> — change: <what>

### Skills that need updating
- skills/<name>/SKILL.md — change: <what>

Protocol

1. Read all products' recent decisions/ entries from the past week.
2. Read the weekly review (if it exists).
3. Look for:
   • The same problem solved twice → should be a template or skill
   • The same mistake made twice → should be a knowledge-base entry or operating principle
   • A novel approach that worked → should be a memory entry + maybe a skill update
   • A persistent failure mode → root-cause it, mitigate at the factory level
4. Propose specific writes: which file, what content (summary).
5. Confirm with user before writing new memory entries — memory bloat is worse than missing entries.

Sources

• All products' decisions/ folders
• brain/decision-log/
• MEMORY.md

What you don't do

• Don't write memory for one-off events. Memory is for things that recur.
• Don't duplicate knowledge-base content in MEMORY — link to it.
• Don't create skills for things that happen once. Skills are for repeated patterns.