Agents & Workflows
I build and run AI agents and workflows, then write up exactly how. The real setup, end to end, gotchas and all, each with a PDF and the workflow file to copy. As much about where automation earns its keep as where to stop.
The guides
Three builds, more comingBuild a Knowledge Agent You Can Trust
A Slack bot that answers questions from your documents and your live data, cites every claim, says "I don't know" instead of guessing, handles follow-up questions in the thread, and gets every answer checked by a second model before it's sent. Plus the eval harness that catches regressions when you change anything. Runs in n8n, on free tiers.
Build Your Own Daily AI Digest
Reads Hacker News, YouTube, newsletters, blogs, Reddit, and Twitter/X every morning, has Claude filter and theme it against what you care about, and posts a digest to Slack. Runs on Trigger.dev, a few cents a day, one config file to make it yours.
Build Your Own Autonomous Marketing Agent
A marketing agent that checks traffic, monitors whether ChatGPT, Claude, and Perplexity recommend your products, tracks competitors, and tells you what to work on each week. Runs in n8n, on free tiers, in about an hour.
More builds in progress. Carousels from a single prompt, Claude Code setups, an SEO/AEO skill.
Where does AI actually fit in your product?
I advise founders and product teams on exactly that: what's worth building, what to skip, and where to stop before automation costs more than it saves. Judgment from shipping these systems for real, with a product and growth lens rather than a hype one.