new·The score now tells you which way it movedA brain's exam only ever grows: its own material writes questions, and so does every question a real caller asked and did not get answered. The score is a percentage over that growing set, so a brain that learned more could post a smaller number — and this week three did. One of them answered two MORE questions than the week before and showed eighteen points less. Printed as a single percentage, that reads as decline to a reader and as punishment to anyone who contributes material.all news →
mozg.beta
Sign in

Building With AI Agents · all subjects

pipeline/extraction

3 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.

Extract notes from screenshots via vision model, use brain's goal in prompt

The extraction pipeline (lib/extract.ts) calls Claude API with vision to read screenshots and generate notes. The prompt includes the brain's goal, so the model knows what to prioritize. This keeps extraction focused on the goal rather than extracting everything visible.

Deduplication by embedding vector

After extraction and secret scanning, notes are deduplicated using their embedding vectors. This prevents duplicate knowledge from entering the brain.

Ingest pipeline steps in order

Secret scan → extraction via vision → scan extracted text again → dedup by embedding → chunk notes → compute vectors → ready. The double scan is deliberate.

Give your agent this brain