Digest

Digest #2: three teardowns and a learn track

Herdr, LifeOS and DeepSeek Harness read against the reference architecture, plus a free seven-module learn track.

Three teardowns published, and a learn track of seven modules.

New public work since the last issue, short notes here, full detail at the links.

Three architecture teardowns are live. The teardown series reads a published agent system against the 18-pattern reference architecture, covering what works, the trade-offs, and what's conspicuously absent. Each is written so the subject could act on it.

A guided learn track, free. learn/ turns the 18 patterns into a seven-module track organised by the six dimensions the maturity check scores, one exercise with a done-check per module. The hands-on companion is the starter template's tutorial, a governed workspace stood up in seven checked steps. If the maturity check told you where your gaps are, this is the lane that closes them.

The Essays section is running. Two published so far. Your agent bill is mostly context, not output is the 147x measurement of what position in the context window does to price. Your agents are about to start working unattended argues that the watching half of unattended operation, the part that checks the work and prices the run, is not being built. Essays argue consequences at the owner's altitude. The digest stays at the workbench.

Same invitation as last issue. If a check is wrong or a teardown misread something, reply. I read everything.

Get new essays by email on Substack. This one is on Substack too.

← All writing