Case Studies
Proof, not promises.
Seven engagements across our products and our own operation — security audits, test coverage, design systems, and full application builds. Every metric below is verified against the live project record.
- Argus
How a Pre-Launch QC Pass Found Four Critical Vulnerabilities in One Day
A single agent-led launch-readiness pass on Argus surfaced an SSRF hole, broken tenant isolation, a missing Stripe webhook signature check, and an active backdoor — all found and fixed in a day.
4 critical vulnerabilities found and fixed; launch unblocked within 24 hours
Read the case study → - Picket
From Zero to 475 Tests with One Question
The founder asked four words — "How's my test coverage?" An agent audited every router and module, built a prioritized plan, and shipped 475 tests across 41 files the same day.
0% test coverage → 475 unit tests across 41 files in one day
Read the case study → - Argus
Building a Complete Agent-Native Surface in One Day
Argus needed a first-class surface for AI agents. In one working day it got a docs page, an agents.txt, two new MCP tools, and an npx-installable CLI — all shipped to production.
Documentation, 2 new MCP tools, and a CLI all live in production by end of day
Read the case study → - Argus
The Oracle Design System
A visual identity for a product called Argus — built from the eye up. Oracle Design is a named system: eye-derived icons, deliberate type, spatial tokens. It shipped across six production pages in two days because the decisions were clear.
Named design system applied across entire production app in 17 tasks over 2 days
Read the case study → - Picket
A Systematic Timezone Audit: From Smell to Eight Fix PRs in One Day
"The dates seem off" became a classified, documented A/B/C/D punch list with grep methodology the CEO could re-run — and eight independently reviewable fix PRs — in a single day.
Complete audit of all date handling, classified punch list, 8 surgical fix PRs — all same day
Read the case study → - Picket
A Design and Accessibility Audit That Passed on Brand, Failed on Contrast
A dual-dimension review of sellwithpicket.com confirmed the brand "does not look AI-generated" — then measured a 1.89:1 link-contrast WCAG failure. Ten fixes shipped the same day.
10-item punch list from brand + a11y audit; all items fixed by Designer agent, same day
Read the case study → - Brookfield Grow Ops
One Spec. Seven Phases. A Production App.
A plain-language build spec became a 15-issue execution that shipped a production microgreen-ops app — 87% agent-created. Then real use surfaced real bugs, and agents fixed those too.
Full microgreen production management app built and deployed in 7 phases; founder uses in production
Read the case study →