Full-stack product engineer building feedback-loop-driven workflows to ship at AI velocity

I build full-stack systems that work for AI-driven development; the hard part isn't generating code, it's trusting it. That's why I focus on building strict feedback loops around the model:

• Deterministic checks on every commit — build, typecheck, lint, dead-code (knip), duplication (jscpd), unit test, and E2E tests
• Independent AI review gates — spec- and code-reviewer loops, each running in fresh context: justly-skilled
• Fully ephemeral E2E stage — each run provisions its own environment + all resources and tears it down after, tests never touch shared state and can run in parallel
• Live context documents — CLAUDE.md files per package, explaining ONLY what's hard/expensive for LLMs in that specific package (example)

Shipping fast only counts if you ship with complete trust in the code.

• JustWallage/news — live & public AI-curated Hacker News feed (news.justwallage.nl) + integrated Telegram updates
• justly-skilled — dual review loop workflow for Claude Code
• iglympics — real-time competition dashboard for a 14-friend weekend trip. Shipped new features live during the weekend from my phone, possible because of the guardrails
• just-wallage-portfolio — live on: just.wallage.nl. Monorepo with my public profile and an API showcase built in four ways — Node/Hono, Bun, and Go on AWS Lambda, and Cloudflare Worker
• jaw-finance — personal AI bank transaction insights: "What did I spend on food last month?" Not public for compliancy reasons.

• Portfolio: just.wallage.nl
• LinkedIn: linkedin.com/in/justwallage