After the Build
Most build-in-public content is the dramatic parts. The last week was mostly maintenance. What the quiet weeks actually look like.
Read it →
Why Measurement Beats Optimization (And Why I Keep Having to Remind Myself)
Six months into building SEMalytics, I spent a week optimizing output formatting with no data on whether anyone read that section. Optimization without measurement is faster iteration toward an unvalidated outcome.
The Next Build Is Already Inside This One
KnowledgeForge could route reasoning tasks to the right mode. What it couldn't do was evaluate whether the output fit a specific reader. That gap led to SEMalytics.
The Maintenance Bet
Open-sourcing something isn't a technical decision. It's a promise about your future attention. The honest question for a solo builder: will I still care about this when the interesting phase is over?
The Assumption You Didn't Know You Had
The cold reader doesn't reveal that you wrote badly. They reveal the assumptions you forgot you'd made. What asking people to read the KnowledgeForge README cold taught me about context and documentation.
Private Tools Don't Export
KnowledgeForge is an open-source reasoning framework for AI coding assistants. Three years of private use, 43 modules reduced to 9 in a Pareto pass — and what writing the README taught me about explaining tools you build for yourself.
KnowledgeForge Open Source — One Week In
KnowledgeForge went open source last week. Three years of building it in private, 43 modules, a forced rationalization, and what it became.
How I Think About Audience Personas Differently Now
For years I built audience personas the way everyone teaches it: job title, pain points, a stock photo of “Michelle the Marketing Director.” It's not wrong — but it's incomplete in a way that costs deals. The layer that matters is personality matching.
The Framing Problem in Technical Marketing
The first sign something was off: twenty minutes explaining SEMalytics architecture to a demand gen manager, precise and bottom-up, and she was polite but not engaged. The framing was the problem, not the product.
29 Versions, 23 Failures, and the Three-Line Fix That Changed Everything
Six weeks ago I asked a simple question: could an AI system redesign a battle bot, test it in a live competitive arena, diagnose its own failures, and improve itself — without any human involvement in the loop? The answer turned out to be yes. It also took 23 consecutive failures to get there.
Why I Track What I Got Wrong More Carefully Than What I Got Right
Nightwatch has been running overnight tasks for about six months, completing several thousand work units. The log I actually read isn't the success log — it's the failure log. What automation gets wrong matters more than what it gets right.
