How I work

AI speeds up the work. It doesn't make the call.

The stack behind this portfolio, and behind most of my day-to-day, organised by where each tool actually earns its place rather than as a logo list. Below is what I use by phase, then the habits that keep a faster process from becoming a less trustworthy one.

The shape of it

The checkpoints between phases are the point. Nothing moves forward on the strength of a plausible-sounding output.

By phase

How I keep control of the outcome

1. 1

   Every claim needs a traceable source.

   If AI-drafted copy states a number or a fact, it has to point back to something real — a document, a screenshot, a transcript — before it ships. If it can't, it gets flagged and either verified or cut.

2. 2

   Unverifiable output gets rejected and rebuilt, not polished.

   AI tools occasionally fill a gap with something plausible-sounding rather than something true. When that happens, the fix is tracing back to what's actually real and starting again from there, not softening the language and keeping it.

3. 3

   One source of truth, not four drifting versions.

   Resume, LinkedIn and portfolio copy are tracked against a single working document, so a phrase or a number doesn't quietly diverge across channels over months of edits.

4. 4

   Technical claims get tested, not taken on faith.

   A tool reporting something as "fixed" isn't the same as it being fixed. Claims like that get verified directly, in the browser, with real interaction, before they're marked done.