Who doesn't need an interactive Ninja Turtle progress bar to make  timeline documents more exciting?

Built simply with Claude to make my colleagues smile. After all, we're all on the same team and building strong relationships is hugely important to me.

Knowing when AI is actually good

The problem: AI-generated translation is easy to demo and hard to trust. Before shipping transcript translation to real learners, we needed a defensible answer to "is it actually good?" — in languages nobody on the product team speaks.

What I did: Designed and ran a structured quality review with native-speaking colleagues and users across Spanish, Finnish, Lithuanian and Romanian AI-generated transcripts. I then synthesised their feedback into a clear ship/fix/hold picture — including surfacing where quality issues clustered and which capabilities stood out (audio-to-translated-text word highlighting proved a standout).

The outcome: A quality translation service for learners with niggles worked out before build. Including colleagues and learners early also built storng relationships and got some fantastic case studies about beingt part of the process too!

Prototype in days, not sprints

The problem: Validating product ideas usually means waiting for design and engineering capacity — weeks of lead time before you learn anything from users.

What I did: I build working, interactive prototypes with Claude Code — real clickable experiences, not static mockups — and put them in front of users and stakeholders while the idea is still fresh. e.g. including early concepts for Genio Present.

The outcome: Ran prototype tests with users within days of an idea surfacing and killed two concepts before they cost engineering time. Aligned stakeholders around a working demo instead of a slide deck.