Software engineer shipping Chrome extensions and developer tools with inspectable proof, narrow privacy boundaries, and honest limits.

Live portfolio · LinkedIn

A dedicated Chrome DevTools panel and npm bridge for registered Zustand stores: live state, a cross-store action timeline, safe in-memory time travel, and optional Pro Trace Sessions with path-level diffs and redacted exports.

TypeScript Zustand 4/5 Chrome MV3 100 automated tests no telemetry

Interactive demo · Source · npm bridge

A local-first workspace that keeps accessibility findings, page context, protected screenshots, retests and client-ready report drafts connected. Capture happens only after a toolbar click; project data stays in the browser.

React TypeScript IndexedDB 77 extension tests 15 website tests

Product site and live demo

Break a deterministic five-node distributed system and watch routing, quorum, latency, dropped writes, and recovery reconcile the same state.

TypeScript React SVG deterministic simulation

Signal Garden Turn microphone or local audio into a reproducible generative organism. The signal never leaves the browser. Web Audio Canvas local-only

Second Order Change one assumption and trace 10,000 seeded futures through a visible causal model. Monte Carlo Web Worker shareable state

The July 2026 launch check for these three systems produced:

• 3 successful production builds
• 16 passing automated tests
• 0 known npm vulnerabilities
• no accounts, analytics, telemetry, remote scripts, or public source maps

Each repository includes its architecture, security model, tests, live deployment, and honest limitations.

The two Chrome products add 192 more automated checks across their extension and website suites. Test counts are evidence of covered cases, not a claim that defects are impossible; I have published the failures that earlier green tests missed.

• Verify — verify a genuine 2048-bit RSA/DKIM signature in the browser without a crypto library.
• Auditable CV — a CV whose public claims check themselves against live GitHub and npm data.

1. Write down the behaviour and constraints.
2. Separate the domain model from the interface.
3. Build the interaction that makes the hard part visible.
4. Test deterministically, audit dependencies, and inspect the production build.
5. State what the system cannot prove.

I use AI coding agents as implementation tooling. Product judgment, architecture, testing, security review, and final verification are mine.

© 2026 Kuba Opoczka.