03 · strategy through production code
Full-stack build
Next.js applications, WordPress estates, integrations between stubborn systems. Nineteen years of shipping means the architecture conversation and the pull request come from the same person.
four products engineered end to end and shipped — and the site you're reading, webgpu field included · scroll — the instrument is live ↓
the instrument — commits become interface — the build, replayed
abstracted from this site's own history
commits becoming interface
01<Header wordmark nav />
02<Headline display chrome />
03<Doors count={3} magnetic />
04<Metrics source="live" />
05deploy — vitals green ✓
✓ deployed — lcp under budget
the work — manifest
w-01
Modern web builds — Next.js, React, edge deployment, the vitals green from the first commit.
w-02
WordPress at production grade: bespoke themes, hardened plugins, WooCommerce, hosting ops.
w-03
SaaS engineered end to end — four products built and shipped, spec to store.
w-04
Integrations and internal tools: CRMs, feeds, dashboards, the glue nobody else wants to own.
w-05
Rescue work — inherited codebases stabilised, documented, and made shippable again.
how it's different
You're reading the portfolio piece. A 110,000-particle WebGPU simulation runs behind this text at 60fps — and the page still loads instantly.
No agency layer cake — the person you scope with is the person who ships.
Repos are yours from day one; so are the keys, the infra, and the docs.
method — how it runs
01
scope against reality
the estimate comes from the person who'll write the code — which is why it survives contact with the codebase.
02
phase with gates
fixed-scope phases, each ending in something running that you can reject. No six-month reveals.
03
instrument as it ships
vitals, error tracking, and measurement land with the feature, not after the incident.
04
hand over like an adult
documentation a stranger can operate from, repos in your name, and no hostage infrastructure.
manifest — what lands on your desk
jm/full·03
- 01
production applications — deployed, monitored, documented
- 02
codebases a second engineer can join without archaeology
- 03
handover docs written to be executed, not filed
- 04
your repos, your accounts, your keys — from day one
signed — one operator, no handoffs
04 items
engagement — the terms
Fixed-scope builds with phased gates, or retained engineering by the month. Repos are yours from day one.
proof — on the record
built & shipped
on recordKlarvo, LessonLoop, SEOflare, Agent-Ready Grader — four products engineered end to end.
the registry →this site
liveNext.js 16, a GPGPU particle field, live AI surfaces, and instruments on every page — one person's commits.
asked straight — answered straight
01What stacks do you actually build in?
WordPress estates through headless and Next.js builds, with the unglamorous layers included: hosting operations, caching and CDN strategy, measurement stacks. The site you're reading is the reference — Next.js 16 with a WebGPU particle field, graded live by its own instrument.
02We've had audits. Can you just fix what they found?
That's the entire point of the practice: audit → commit, not audit → backlog. Send the findings and I'll tell you honestly which are worth fixing, which are theatre, and then ship the ones that matter.
03What does a performance rescue look like?
Root cause, not symptom: server response, render paths, caching layers, delivery. The one number this site publishes is from exactly that kind of engagement — a UK tile retailer's server response taken from 2.2 seconds to 47 milliseconds, measured either side of the fix. Every other figure stays off the record until it's verified.
04Can you rebuild a site without losing its search equity?
Yes — by making it boring. The URL inventory is frozen before a line is written, the redirect layer is generated and tested at runtime rather than hoped about, and rankings are watched through the flip. This site's own rebuild carried 249 URLs across with every redirect verified single-hop.