How I work

Hands-on in the product, and across the organisation.

AI for the whole organisation

As a tech lead, my responsibility is to translate user problems into technical solutions, and I'm hands-on building them with the team. The harder part is not the building: getting them out to end users needs to run through product teams, sales, marketing and support as well. Getting them right for the users is the discovery work below.

Helping colleagues use AI in their own work is a separate part of the job. Much of it is about helping people who don't write code get real value from AI: Claude workshops with managers and with the legal team, and non-developers working in GitHub alongside developers, editing with AI help, under the same version control and review as the code.

From problem to production

Discovery before delivery, small teams that get from problem to production fast, and stopping what the evidence doesn't support, as with Kompass.

I run my own development through AI agents with defined roles. I own the scope and the judgement calls.

• Every agent gets a written brief, and reports back what it concluded, not everything it read.
• Reviews are independent. A reviewer never saw the conversation behind the work, so it doesn't inherit its blind spots.

The setup is public: the skills and role briefs are in my dotfiles.

One round from problem to production. The agents do the work; nothing goes out without my go.