Engineering
Notes on system design, the decisions behind the product, and lessons from operating an AI-managed marketing platform for small businesses. Opinionated, occasionally wrong, always honest about tradeoffs.
sitemap.xml tells search engines where your pages are. llms.txt tells AI assistants what your business actually is. Every site we host now serves one — generated, deployed, and kept current automatically.
Six weeks of real meter readings from running autonomous agents for small businesses — what a run costs, the night retries burned $11.75 while everyone slept, and the five cost controls that came out of it.
For most SMB-facing AI products, running your own thin agent loop against an LLM API is the right call. Here's when the platforms make sense and when they're just expensive indirection.
A one-line install, a sandboxed runtime, and a security model that survives someone copying your snippet. The decisions behind an embeddable AI assistant.
Default LLM voice is too hedgy, too disclaimer-heavy, and too transparent about how it's answering. Five small prompt rules that swing tone hard.
We ship new posts roughly monthly. Message the bot on Telegram and ask it to nudge you when something new drops — it can do that.