WritingReading path
Reproducible recipes from systems running here — editorial agents, retrieval stacks, conversion plumbing, and a live AdCP signals agent. What shipped, and what broke.
8 essays in this path
of 94 published · 2025–2026
A complete build recipe: GA4, Search Console, Claude, and GitHub wired into an editorial agent that reads a month of data, writes five ranked recommendations, grades its own previous advice, and pings you — on two cadences, for pennies. Five APIs, zero dashboards, one afternoon.
Dashboards don’t fail because the charts are wrong — they fail because a chart is a question and you were promised an answer.
I rebuilt the retrieval system behind Ask Evgeny in one day: traces on every query, a 975-node knowledge graph, authority buckets on every position, and three model upgrades running in shadow. The most important thing that shipped is the feature that measured itself and refused to launch.
The signal architecture that makes one conversion land once across Google, Meta, and OpenAI ads — why the three platforms are complementary, why you can't dedupe walled gardens but can out-hub them, and the incrementality question underneath all of it.
A complete build recipe for putting a grounded answer engine on top of your own writing: chunk the corpus, embed it, retrieve with a multi-lane hybrid, generate with hard grounding rules, track when your own positions change so the system never quotes a retracted stance as current, and expose it two ways — as an MCP connector any chat client can use, and as a chat box on your own site. Every part runs on one Cloudflare Worker for near-nothing. The honest version: the architecture, the guardrails that keep it from lying, the four bugs that actually cost me an afternoon, and the loop that makes it smarter every time you publish.
What running a live AdCP signals agent teaches you that reading the spec never will. Revised September 2026: a privacy gap I filed against Trusted Match shipped in AdCP v3.1.22 — on stricter terms than I proposed, in a maintainer PR rather than mine — and two threads listed here as open proposals have since closed as completed.
A reproducible Rube + Claude + Slack recipe. From inbox to scored lead in under 30 seconds.
Start building it. A bi-monthly tutorial series on shipping real GTM agents in Rube, Claude, n8n.
A walk-through of anonymized PII + psychographic profiling + LLM-driven creative for a single high-value persona.
1 appearance
From Data to Impact: How Martech Drives Real Business Growth
Watch the panelAsk about the essays, frameworks, playbooks, or how to work with Evgeny.