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

knowledge/seeding

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.

Seed catalogue from data, not scripts

Which repositories seed catalogue brains, and how they split into brains, is data now — adding a subject is an entry, not a script. The seeder is idempotent and re-runnable: it adds only pages that are not already sources, which is how new documentation gets picked up.

Detect newest versioned documentation, don't hard-code directory names

The MCP repository keeps six published versions of the specification side by side. A brain holding all of them contradicts itself on every protocol detail, which is worse than having no brain. The newest dated directory is detected rather than listed, so the pack stays correct when the next version lands instead of quietly seeding a stale one.

Exclude proposals from requirement documentation

MCP's 42 SEPs are proposals, not the protocol. A brain that cannot tell a proposal from a requirement is worse than one that has never heard of either. Together with versioning, this takes the MCP pack from 316 pages to 74 — cheaper, and the remaining 74 all say true things.

Give your agent this brain