Writing & build logs.
Essays on software development, web builds, and AI tooling. Plus build notes from what I'm actually working on right now.
Copy bugs don't throw exceptions
The NAPA SEASON album page attributed production and vocals to Suno. That gave authorship to a rendering engine. The /contact timelines note filtered out prospects before they had sent a message. Neither bug showed up in any log.
The calculator passed every technical check. Three AdSense rejections later, the hub pages were the problem.
Three identical 'Low value content' rejections, all with empty policyIssues fields and an authorised ads.txt. A 15-agent root-cause review pointed at the content estate, not the calculator logic. The fix was architectural, not compliance-oriented.
The timelines note that was filtering out the right prospects
The /contact/ page carried a note about timelines, written to set expectations and catch poor-fit enquiries before they became calls. Re-read through a prospect’s frame, it wasn’t catching poor-fit enquiries — it was rejecting good ones.
The work history cards shipped. Two were wrong within the hour.
Four cards went into the web history section of `/work/`. Two needed correcting before the hour was up — one misattributed, one carrying a name that had been retired six months earlier. Both failures had the same root cause.
Five copy edits on captainrandom this quarter. None were about wording.
A blurb gave Suno the authorship of an album I wrote. A sub-brand name attributed the creative work to the AI. A contact note read as a rejection of prospects, not a quality signal. Five copy edits in two months, and every one corrected a fact about authorship or positioning — not a sentence.