The album page for NAPA SEASON had one sentence that was wrong in a specific way. The body prose read “Suno for the production and vocals.” Suno is a tool. The lyrics were written here. The melodies were built in Logic. The production decisions were made here; Suno and the surrounding toolchain handled the rendering of vocals and instruments. The distinction — between “the tool executed” and “the author decided” — is what separates “AI-generated music” from “music made using AI tools.” The blurb had it backwards. It was giving away the authorship of the record.
That matters on a creative portfolio in a specific way. The claim this site is making is that there’s a person here who uses AI in the production chain, not an AI pipeline that occasionally receives human input. One sentence had erased that distinction.
Fixing it was one sentence. Looking at the surrounding pages afterwards, four more edits had been waiting for the same reason.
The /creative/ corrections
Three of the five edits were on the creative section of the site.
The NAPA SEASON fix was the most direct: wrong attribution, corrected. The sub-brand rename came from the same impulse. The music production work on the site had been labelled “Creative AI Series.” That name frames the output as AI’s work with a creative modifier applied. It isn’t. The rename to Creative AI Assisted Series puts the human role where it belongs in the name itself — the series is made here, AI assists in that process. Seven places carried the old name: single pages, album pages, the footer, the videos and series index descriptions, the SingleRelease schema component. All updated in the same pass.
The Tasogare edit came earlier. Tasogare — Cinematic Boom Bap was the first track, uploaded to YouTube and featured on /creative/ to replace the “first track lands soon” placeholder. That commit also corrected the DAW lineage — the production chain for the work on this site is specific, and the page now reflects it accurately. Same pattern: the original copy had left the authorship of the technical process ambiguous. The correction made it precise.
What the /learning/ hub said the catalogue was
The /learning/ hub wasn’t misattributing credit to a tool. The problem was that the heading wasn’t claiming anything.
A catalogue heading that describes content type tells a visitor what they’re browsing. It doesn’t tell them what the person who built the catalogue thinks it’s for. The hub h1 is now “Engineering thinking.” The sub-heading, the meta description, and the hub JSON-LD description carry the same framing: courses that teach how an engineer reasons — the decisions and trade-offs, not just the mechanics.
That is a position. It’s also an authorship signal: this catalogue was assembled by someone who has a view on what engineering education should be, not someone compiling a reference list. The old heading declined to make that claim. The new one makes it.
No new content was added. The same courses were already there. The edit replaced a neutral description with a specific one that said what the author thinks the collection is.
What the /contact/ note said about whose time mattered
The timelines section on /contact/ had been written from the wrong point of view.
The old copy ran roughly: rushed engagements aren’t being taken on; two weeks is unlikely to be enough runway; if that’s the constraint, this is probably not the right fit. Read from behind the keyboard it describes a standard — I don’t cut corners. Read from in front of it, by a prospect with a real deadline, it reads as a rejection delivered before they’ve made their case.
The rewrite kept the same constraint. Rushed work is still not taken on. What changed was the frame: the copy now presents the constraint as a consequence of how the work is done, not a door being closed on the reader. The underlying fact didn’t change. The audience the paragraph was addressing did.
The pattern
Five edits across six weeks. None were part of a planned content audit. Each surfaced when a page was read through a frame that wasn’t the author’s own — a collaborator’s, a first-time visitor’s, a prospect’s.
The common thread is authorship. Not voice consistency, not keyword coverage, not information architecture. Who made what, whether the record reflects that, and whose interests the text is serving.
A portfolio is a set of claims. Each sentence either stakes a position or declines to. “Suno for the production and vocals” stakes a position; it’s wrong. “Creative AI Series” declines to name the human role; the omission misleads. A catalogue heading that describes content type without positioning stakes no claim about what the catalogue is for. A contact note written from the author’s perspective rather than the prospect’s is serving the wrong reader.
None of these were breaking anything. The pages were functional. The site was getting traffic. No complaint arrived about the NAPA SEASON attribution.
But a portfolio that misrepresents who made what does quiet damage over time. The wrong inference gets made. The right reader arrives at a conclusion that isn’t true and leaves on that basis. The contact note screens out prospects who would have been a good fit if they’d read something that described the offer accurately.
Catching these required reading each page as someone who didn’t already know what was true. That’s harder than a typecheck. There is no lint rule for “this sentence credits the wrong entity.” No build step catches “this heading is neutral when it should be specific.” No automated test flags “this paragraph is addressed to the wrong person.” The only method is re-reading from the outside, which produces fewer errors per pass than writing from the inside.
The fixes are in the build log. A site that uses AI tools as part of its output — music, copy, code — is going to keep producing authorship ambiguity at the edges. The method stays the same each time: read from the outside, find where the record diverges from reality, correct it.
