I find where software and AI answers break, write it up so it gets fixed, and help the person who hit it.

Promo codesBlocked Builder7 steps86.0sShips SAVE10 and FREESHIP Red team11 steps58.0sRows do not add up Judge3 steps51.7sBlocked, citing e12, e13, e5 Tests pass Builder ships Red team attacks Bug found Blocked QA gate A builder ships promo codes, a red team attacks the change, and a judge blocks or approves it, citing the evidence.
dog barking cat 1 cat 2 cat 3 dog 1 dog 2 dog 3 birds 1 birds 2 rooster cow baby sneeze rain 1 rain 2 thunder 1 thunder 2 30 clipsSearched by sound, in your browser Audio retrieval eval Search 30 clips by how they sound, not their file names. Add up to 10 of your own; they stay in your browser.
Self-healing test A UI change breaks a browser test. An agent reads the failure, patches the selector, and the test reruns green.
Your cart Notebook1$12.50 Pen set1$8.75 Tote bag1$24.00 Subtotal$45.25 Shipping$5.00 Total$50.25 Free shipping on orders of $50.00 or more. Place order IssueNegative quantity accepted: subtotal drops and order can be placed Pull requestClamp typed cart quantity to at least 1 CI passes8 passed, 0 failed Bug to fix A tester files an issue, an agent reproduces it with a failing test and opens a pull request, and the CI check passes.
Adobro Tap a mouse button, click any panel on screen, and a short AI answer opens beside it, looking things up on the web when it needs to. A Windows tool I built and use.

1 of 5

About

I'm a software QA engineer in Toronto, and for more than six years I've tested a professional desktop app as an SDET at Audiokinetic, part of Sony Interactive Entertainment. I have a diploma in sound engineering from Nimbus Recording Arts, and I also work part time in live operations at Keywords Studios. On the side I build agent workflows and evals, which is most of what you'll find here. Outside work I sing, hike, play FPS games and JRPGs, and hang out with my dog.

Method

Agentic engineering, to me: verification is the most important part, then encoding what I learn from an agent's failure modes, because they're very teachable.

How I run a project

$ idea --to ship

1ResearchYou and agents explore the idea with the right playbooks and skills.
2GateA gating agent proves it's feasible, or it gets scrapped.
3BriefThe chief of staff (CoS) gets briefed and oversees it across projects.
4SpecPMs draft and lock the spec, and verify the needed skills are built before any code.
5BuildEngineer and QA agents work with evidence, not claims.
6AttackRed teams try to break it, and failures loop back.
7Ship gatepre-ship-check must pass, and failures loop back.
8LearnFailures and lessons feed back into the playbooks and skills.

It's exactly how I set up my multi-bot harness.

Toolkit

Verification and shipping

pre-ship-check: The ship gate. No proof, no ship.
create-verification-skill, maintain-verification-skill: Give agents hands and eyes for each app.
verify-portfolio: This site's own proof run.
webapp-testing: Drives the browser to test web apps.
tech-debt: Finds what to clean up before it gets expensive.

Build and review

architect: Settles the types and structure first.
arena: Runs parallel attempts and keeps the best.
interrogate: Adversarial review across models.
blast-radius: Shows what a change could break.
benchmark-checklist: Measures before claiming something got faster.

Design and UI

impeccable: A design rulebook on top of a shared design system.
audit, critique, polish, layout, adapt, colorize, bolder: Focused passes over a screen, one concern at a time.