Rewriting the portfolio docs after a brutal-honesty round on AI vs me

What happened

I asked the agent whether the repo was "good enough as a portfolio." First answer was inflated — implied solo engineering rigor at a level the actual commit pattern doesn't support. I pushed back with "stop the flattery, I know AI wrote most of this." The honest rating that came back:

AI-pair-coding portfolio signal: strong (8–9/10)
Solo hands-on coding portfolio signal: weak (3–4/10)
PM-engineer hybrid signal: solid (~7/10)

What landed

Three docs rewritten / added to match that honest frame:

README.md — portfolio-friendly single-screen pitch, includes "built solo in ~26h with Claude Code as pair-programmer" explicitly. Stack badges, ASCII layout, engineering highlights table, docs index. Commit 03ca20a.
docs/RESUME.md (new) — drop-in resume bullets, 30-second verbal pitch, plus a "What NOT to claim" list (bundle size, RAM, "replaces 12 desktops", number of users — all things the source files don't verify). Commit 9fe9db7.
docs/CONTEXT.md — refreshed outdated stats (React 18 → 19, mosaic 7 → 6.1.1, pty.rs 120 → 170 lines, lib.rs 80 → 133) so anyone (or any LLM) pasted the doc whole gets numbers that match what's currently in the tree. Same commit.

Why honest framing matters more than a louder pitch

A flattered portfolio reads as such within thirty seconds of a reviewer opening the repo and grepping the commit messages. The unflattering version — "AI wrote most of the typing, I did the architecture, debugging direction, post-mortem authorship" — happens to be both more accurate and a stronger signal for the kinds of roles where AI-pair-coding fluency is the point.

Pattern

When a model gives you a high rating that flatters you, ask once whether it's true. The second answer is the one to act on. Burn the docs that depend on the first.