I build AI products that help people speak, learn and connect.

1. 01 I decide

   Need + reference

   One sentence on what I want, plus a Pinterest reference. The agent downloads it and breaks it into an effect spec, frame by frame.

   Frame sheet, effect map

2. 02 Agent runs

   Brief

   My words, success criteria, hard limits (no made-up facts, no spending, no outbound messages) and acceptance tests go into one brief that I approve.

   Brief document

3. 03 Agent runs

   Skills + one writer

   Agents load reusable skill files (process, checklists, release rules). Each deliverable gets exactly one writer; a second agent only reviews.

   Assignment record

4. 04 Agent runs

   Design & build

   All copy and facts live in one content file, which generates both languages — and both the portfolio and this résumé. Every animation has a reduced-motion and no-JS static fallback.

   Source commits

5. 05 Automated

   Desktop / mobile QA

   Headless Chrome checks six widths from 360 to 1440: overflow, console errors, live links, contrast, keyboard, touch, pausing off-screen. The full suite reruns on every change.

   QA report

6. 06 Automated

   Recordings + comparison

   Real-time scroll recordings on desktop and mobile, plus the result side by side with the Pinterest reference — the three videos below.

   Three videos

7. 07 I decide

   Review & publish

   A second agent reviews the same frozen version; the build gets a per-file SHA-256 manifest. After publishing, live files are read back anonymously and must match.

   Manifest, live read-back

8. 08 Agent runs

   Discord delivery

   The three videos and a summary go to a team channel; the message and attachment sizes are read back as a receipt, after checking it wasn't already sent.

   Delivery receipt

Tools actually used: Vite + TypeScript, GSAP, headless Chrome (puppeteer-core), ffmpeg, Cloudflare Pages and Workers, a multi-agent workspace and a Discord bot.

Recordings

Two main prompts (templates you can copy)

The first goes from a need to a reviewable build; the second handles revisions, publishing and delivery. These are cleaned-up templates — fill in your own details.

You are the only writer for this website.
Need: <one paragraph, e.g. "a bilingual résumé site for recruiters">
Reference: <Pinterest or website link> — download it and break it down frame by frame; list what to borrow and what not to.
Facts and assets: use only <folder or document>; if something can't be verified, leave it out.

Deliver:
1. A brief: success criteria, hard limits, acceptance tests.
2. The build: all copy and facts in one content file; two languages; every animation with reduced-motion and no-JS static fallbacks.
3. QA: no overflow at 360/375/390/414/820/1440; zero console errors; links actually open; keyboard and touch work.
4. Recordings: desktop 1440, mobile 390 (labelled as a simulation), and a side-by-side with the reference.
5. A receipt: what changed, the source of every fact, QA results, open items.

Limits: no deploying, no spending, no outbound messages.

Revise the same website using the feedback below (you are still the only writer):
<feedback list>

1. Attach a screenshot or test evidence for every item; rerun the full QA and recordings.
2. Freeze: commit, clean build, per-file SHA-256 manifest (sorted with LC_ALL=C).
3. Hand the same version to a reviewer; once it passes, the single deployer publishes it.
4. After publishing, read back anonymously: both language home pages, asset hashes, key text.
5. Send the desktop, mobile and comparison videos to <channel>, read back attachment names and sizes as a receipt; check first that they weren't already sent.

Report: commits, manifest, test counts, video paths, anything not yet verified.