I'm an interpreter, not a job title. I feel what's missing in a system and build the fix. Every project I've done lowers the resistance between a person and something they could already reach — musical thought → sound, information → knowledge, obscure research → the public. The current is already there; I remove what's in its way.
One motion, 30 years, many domains: I find a pattern of manual labour, organise it into a principle, and the principle becomes a tool. The domain changes — music, code, archive, browser — the principle doesn't. A BBS in 1992 and an agentic AI browser in 2026 are the same move, 34 years apart.
Most teams split testing in two: a tester who files tickets, a developer who fixes them, a wall between. I'm the join — I test the code, find the exact lines, write the fix, verify it, and catch the usability nobody assigned me. One person who closes the loop is worth more than a separate tester and developer, because it never has to cross a desk.
I sit between designer, developer, and AI. I write the tests for how a feature should behave before the code exists — spec-driven, test-driven, increasingly agentic. A clear spec up front is the most valuable thing in the room.
I close the gap, I don't just report it. I build the fix, verify it, hand over a confirmed patch — including the small stuff a product owner de-prioritises to avoid interrupting a senior. Seniors stay on the hard problems; the backlog still ships.
I catch it while the developer's still online. A crash-on-launch lands at 15:59 — a two-minute walkthrough now beats a ticket they'll context-switch back into next week.
I learn by watching people use it. Live demos to prospective buyers; PlayTestCloud recordings and transcripts. A spec says what should happen; a person's hands say what does.
I test the organisation, not just the product. A thousand green tests don't prove a thing is good. Where's the bottleneck, the bus-factor-one, the onboarding too weak to make a new hire productive? That's the difference between shipping quality and shipping activity.
Red-pen psychosis. If 150 tests pass but it's visually broken or takes 17 clicks to do one job, it isn't done. On a streaming service I caught four-year-olds' profiles with access to ultra-violent titles — broken age-rating ingestion everyone else had signed off as "looks good." I looked at it as a parent.
Generative AI — I understand it from both sides
As a tool. I work in it daily (Claude, Gemini, Cursor, Codex). At my last role I built my own AI development assistant — a local helper that learned the product's architecture and conventions from the team's own commits, then could take a screenshot of a wrong icon, fix it, and return a verified build.
As a product. I can appraise how a product uses AI — whether it's genuine intelligence that helps the user, or a thin wrapper that looks clever in a demo and frustrates in daily use.
Why this matters to whoever hires me
I live and breathe the product — not something I switch off at the door. Give me a small, agile team and the autonomy for deep, focused work, and I close gaps nobody else was looking at.
The drop:Not a tester in a box filing tickets — product intelligence. I sit between designer, developer, and AI and close the gap until the product actually works and feels right.
Appendix — the fuller passages
Nothing below is new. It's the longer wording that was condensed in the sections above, kept here so no detail is lost.
The credo, as a line: I identify a pattern of manual labour, organise it into an underlying principle, and that principle becomes a tool.
The interpreter motion: the work looks cross-domain from the outside, but it's one motion from the inside.
Closing the gap, in full: I build the simple fix locally, verify it works, and hand a senior developer a confirmed patch — because a senior's time belongs on high-impact architecture, not on a typo. I'll happily "cosplay" a junior dev to make the real fix land faster.
The 4-o'clock silo: a build lands at 15:59 with an obvious crash-on-launch; if gloves come off at four and nobody looks until nine the next morning, that's a bottleneck. A two-minute sanity check while the developer is still online means a working build by morning.
The AI assistant, in full: a local helper that learned the product's architecture and code conventions from the team's own commits and how they talked about them, then could take a designer's screenshot of a wrong icon, fix it, and turn around a verified build. I optimise every bottleneck I find, on both sides of it.
Watching people use it, in full: at my remote-support role I ran product demos for prospective buyers — on-site, over the internet, and at the office — and watched exactly where real people stumbled. On a mobile game I studied PlayTestCloud recordings and, especially, the video transcripts for the same reason. A spec tells you what should happen; a person's hands tell you what does.
Red-pen psychosis, in full: I can't read a restaurant menu without seeing every stray comma and misalignment — and an app is the same. If 150 automated tests pass but the thing is visually broken or takes 17 clicks to do one job, it isn't done. I push back on that, constructively, and offer the alternative.
Why this matters, in full: I integrate into a company and live and breathe the product — it isn't a thing I switch on when the office door opens and off when it closes. I work best with the autonomy to find the place and time where I do my sharpest work, deeply focused rather than in an open-plan room with seventeen conversations at once. Give me a small, agile team where the role isn't "test it, list it, throw it back over the wall," and I close gaps nobody else was looking at.