I re-read the /contact/ “Note on timelines” through a prospect’s eyes. The note said, in substance: not taking on rushed work, two weeks of lead time required, if your deadline is tight I’m probably not the right person. I wrote that as a quality signal. A prospect reads it as a rejection notice before they’ve even typed their name.
Both readings are accurate. That’s the problem.
The July copy audit surfaced that failure mode five separate times across four pages. Different surfaces, different contexts, same root cause: wording accurate to the author and misleading to the reader.
The failure mode
It isn’t a factual error. A factual error is wrong for everyone — the number doesn’t match, the link is dead, the claim isn’t real. You catch it by checking the source.
This is different. The copy is accurate. The author reads it and it says exactly what they mean. The gap only opens when someone else reads it — and they’re not asking “what did the author intend?” They’re asking “what does this tell me?”
The verification reflex that catches factual errors doesn’t catch this. You need a different check: read from the reader’s position, not your own.
Five examples
The timelines note. The /contact/ note existed to set expectations. Two weeks of lead time, no rushed work, real constraints stated upfront. Written to establish credibility and filter bad-fit enquiries. What a prospect sees: a paragraph of reasons to click away. The fix wasn’t to remove the constraint — the two-week lead time is still there. It was to stop doing the prospect’s rejection for them. “Probably not the right person” is a reader delivering their own verdict before they’ve sent a single line. The same facts, repositioned as a statement of how good work runs, read differently.
The /learning/ hub heading. The hub had a heading that described the catalogue accurately — accurate to what was in it, not to what it’s for. The new h1 is “Engineering thinking.” — carried through the sub-heading, the meta description, and the JSON-LD. The catalogue is courses that teach how an engineer reasons: the decisions, the tradeoffs, the mental model behind the syntax. The old heading described the contents. The new one describes the value. The catalogue didn’t change; what a first-time visitor understands about it did.
The sub-brand rename. Seven spots across the site carried the label “Creative AI Series.” Accurate to the production process: the music is AI-assisted. What it communicates to someone reading without context: AI-generated music, with all the connotations that phrase carries in 2026. The rename — “Creative AI Assisted Series” — is one word longer and substantially more precise. The authorship is human; the tools are in the chain. One word does that disambiguation across all seven spots.
The NAPA SEASON blurb. The album page had a line attributing production and vocals to Suno. Accurate to the render step — Suno is at the end of the production chain. Not accurate to authorship: the lyrics and production are written here, melodies built in Logic, Suno handling the final audio generation. The blurb read “Suno made this.” That isn’t a framing problem; it’s a factual misattribution of creative work. The correction named where Suno actually sits in the chain and restored the authorship record.
Tasogare. This one runs differently. The /creative/ page held a “first track lands soon” placeholder where the Tasogare feature card now sits. The placeholder was accurate when written. Once Tasogare — Cinematic Boom Bap landed on YouTube, it was accurate to nothing — stale, not misleading in the author-vs-reader sense. Both the author and a reader would read “coming soon” and know it was wrong. The fix was replacing the placeholder with the actual card and the live link.
What four of the five share
The timelines note, the /learning/ heading, the sub-brand label, the NAPA blurb: all four were written from the author’s position and published without asking what they communicate to someone who doesn’t share that position.
I know the timelines note is a quality signal; the reader sees a discouragement. I know the learning catalogue is about engineering reasoning; the old heading didn’t convey that to someone arriving cold. I know “Creative AI Series” means assisted, not authored; the label didn’t carry that distinction. I know Suno is a render tool, not the author; the blurb didn’t say so.
In each case the author and reader were reading the same words from different positions. The author position was checked; the reader position wasn’t.
Why it’s harder than a factual error
A factual error has a source to verify against. The number either matches or it doesn’t. The workflow is direct.
An author-vs-reader gap requires holding two simultaneous readings of the same sentence: what you intended, and what the words actually do when they arrive in a different frame. That’s not verification. It’s closer to reading as if you’ve never written it — which is a harder state to get into when you have written it.
The check I’ve been running: treat the copy as a question. What does this sentence tell someone who doesn’t already know what I know? If the answer is “something different from what I intended,” the sentence is the bug. If the answer is “nothing,” it’s a placeholder or dead weight.
What this isn’t
Not an argument for softer copy. “Engineering thinking.” is more specific than the old heading. “Creative AI Assisted Series” is more precise. Two weeks of lead time stated as a standard is still two weeks of lead time. The constraint isn’t removed in any of these fixes; the framing shifts.
The goal isn’t to soften claims. It’s to make them communicate what they actually mean to someone reading without your context.
The log
All five are in docs/devlog.md. Tasogare 2026-06-22. NAPA SEASON correction 2026-07-14. Sub-brand rename 2026-07-17. /learning/ heading 2026-07-21. /contact/ timelines 2026-07-25. One paragraph each: what changed, what the old version communicated, what prompted the re-read.
Five fixes. One bug.
