Quezzies — Engineering

Your app works in the demo. I make it work in production.

The Three-Day Launch-Blocker Sprint: I take one customer-, launch- or revenue-blocking problem in an existing TypeScript or AI-built application and get it working in production within 3 working days. Implementation, focused tests, deployment, written handover, and 7 days of defect correction. One repository, one production outcome. Fixed £3,000, paid upfront to reserve the sprint.

Book a 15-minute triage

2 slots per month · run alongside my own venture · yes/no on the call

Sound familiar?

A customer can't onboard because authentication is broken.
Users can see one another's data.
Stripe takes the payment but never provisions the service.
The production deploy fails, or only works from your laptop.
The AI workflow is too unreliable to demo.
An investor or customer demo is next week and the prototype won't survive it.

You already tried the AI tools — that's how the prototype exists. You're not paying for code an LLM can write. You're paying for an engineer to enter the real codebase, diagnose it, integrate the systems, deploy it, test it, and take responsibility for a pass/fail outcome.

How it works

Day 0

Triage, 15 minutes

You show me the blocker. I say yes or no on the call, and if it's yes we write the acceptance test together — the sentence that defines done. Fixed scope, paid upfront to reserve the slot.

Days 1–3

The sprint

I work in your repository: diagnosis, implementation, focused tests, deployment to production, and a written handover of what changed and why.

After

Seven days of defect correction

If the delivered outcome breaks within seven days, I fix it. No retainer required, no meter running.

“Done” means an acceptance test passes

Every user can only access their own organisation's data — proven with adversarial tests.
Stripe checkout, webhook verification and provisioning work in production.
The application deploys reliably: staging, production secrets, CI.
The AI workflow has structured outputs, failure handling, evaluations and monitoring.
This customer integration works end-to-end and is documented.

If your problem can't be given a concrete acceptance test, I'll say so on the triage call and point you somewhere better suited.

Recent production work

Case // data isolation

Multi-tenant SaaS: cross-tenant data leak, found and sealed

Any signed-in organisation could retrieve another organisation's records — including by guessing direct IDs. Diagnosed the query layer, scoped every read and write to the active organisation, and proved it with adversarial tests: a second tenant attempting enumeration and direct-ID retrieval gets nothing. Shipped through CI to production in two days.

Case // auth + release integrity

Fail-open service auth and unverifiable deploys, hardened

Internal service tokens authorised every request when unset, and build artifacts couldn't be verified after deploy. Made every credential fail closed with constant-time checks, signed the artifact chain so tampered builds are refused, and made renders byte-reproducible. Regression tests pin all of it.

Run by Louis Newman, founder-engineer. I build with the same agentic tooling that produced your prototype — Claude Code, Codex, Cursor-class workflows — every day, at the frontier. Which is exactly how I know where they break.

For agencies

One senior TypeScript / Cloudflare / AI delivery block, available white-label. A contained auth, data-isolation, billing, deployment or AI-reliability problem your core team shouldn't be pulled onto. £1,000/day, 3-day prepaid minimum.

What I don't take

Redesigns, rebrands, or greenfield builds.
“Make it better” without a concrete acceptance test.
Multiple problems in one sprint — one repository, one outcome.
Ongoing unlimited availability. (One capped day-per-week retainer exists for past clients.)

Tell me what's blocked

Two sentences is enough. I'll reply with a yes, a no, or one question — within a working day. Or skip the form and book the triage call.