This Portfolio
The showroom itself: a full-stack Next.js + NestJS app taken from spec to production in under 24 hours of agent-driven work.
Product
The shipped product: live link, the problem it solves, and the stack it runs on.
The problem
Most portfolios show only the finished product. This one shows the build too: the PRD, the commit/PR/deploy timeline, the metrics, and the agent-driven workflow behind each exhibit.
Stack
Screenshots
How I built it
Product requirements
Open the PRDThe full PRD lives in the repo. Open it above.
Built in public by dogfooding the agentic-dev-kit: every milestone ran the PRD → implement → verify → reconcile loop with Claude Code and a custom MCP server. The PRD is public and kept honest against what actually shipped, scope changes logged, not hidden. This showroom is exhibit #1.
Continuous integration
Every push runs the verify gate in GitHub Actions: type-check, lint, and build across the monorepo. This badge is live.
Timeline
- milestoneJun 22, 2026
M3: design system wired (Tailwind v4 + tokens + shadcn)
- milestoneJun 22, 2026
M2: public API + RSC reads
- milestoneJun 22, 2026
M1: backend foundation (NestJS + Prisma + Postgres)
- milestoneJun 22, 2026
M0: monorepo scaffold (pnpm + Turborepo)
Recent commits
All commitsLive from GitHub, as of Aug 8, 2026
- c8d2847Jul 26, 2026
feat(web): blipsmith exhibit, an agent-playable deterministic synthesizer (#27)
- 0264d62Jul 15, 2026
feat(web): CV positioning to Full Stack Developer + SE fundamentals (#26)
- b44f715Jul 8, 2026
feat(web): reposition CV around generative AI + per-role tech lines (#25)
- e326167Jul 8, 2026
feat(web): refresh CV with metrics-led positioning + new projects (#24)
- c3a9758Jul 7, 2026
feat(agent): live exhibit screenshot + B11 deploy runbook and conventions (#23)
- 356ae6bJul 7, 2026
feat(agent): Portfolio Agent, a RAG concierge built in public (#22)
- 50fd69dJul 7, 2026
- 1fc861eJul 2, 2026
feat(web): positioning, conversion, and detail-page overhaul
Metrics
Designed, built, and shipped in under 24 hours for under $15 of agent time, about 15% of a weekly Claude Code plan. That is roughly 7 projects of capacity in a five-day work week. Capacity, not yet a track record: the cadence metric stays a target until it is proven.
Each metric carries an honesty tag. Verified numbers read solid; targets and placeholders read muted.
Built the same way, every time.
Every exhibit runs the same loop: spec, implement, verify, ship. If you want this kind of work on your team, full-time or freelance, let's talk.