Browse all posts by topic, or start with a learning pathway.
Showing all 23 posts
The first time I ran an AI-assisted triage system properly, it burned through 5.2 million tokens and told me almost nothing useful. Six weeks later, running it again, I burned through nearly as many tokens a second time, for a completely different reason.
Six weeks touring every department before touching a single system, a work trial that starts at six in the morning with coffee waiting: what onboarding is actually for, and what happens when it's left to assemble itself.
Hiring is the upstream part of building a high-performing team — the conditions you create before anyone joins. A system, not a transaction.
How AI fills the gap that teams provide for solo practitioners—holding execution work so thinking can happen continuously
When work is invisible, people fill the gaps with assumptions. They assume they're failing, that others don't care, that the work doesn't matter. You can't fix morale or trust or delivery by addressing those symptoms directly. You have to make the work visible.
What dysfunction feels like from the ground floor — a story about being told off for helping people
How I helped turn around a team with 54% attrition and -37% eNPS by rebuilding connections, creating ownership through participatory OKRs, and reversing a negative spiral into positive momentum.
Moving beyond individual documentation practices to building team-wide knowledge ownership and teaching culture
Leading a £10M capital investment to build automated PPE manufacturing during COVID—proving UK efficiency could compete with Far East costs, then watching the government offshore it anyway.
Five dependency upgrades taught me the difference between AI iterating normally and AI guessing wildly. Knowing when to intervene makes all the difference.
How we spent three years systematically replacing our technical architecture whilst delivering continuous business value - not in a big-bang replatforming, but plank by plank, always sailing, always improving.
A 2x2 grid that explains why some working relationships work and others don't. Care personally and challenge directly - but most of us drift between all four quadrants depending on context and pressure.
The two modes in practice - 90 minutes from exploratory thinking to live deployment, pushing into unfamiliar territory with API keys and GitHub Secrets.
From local development to production in a day - learning Azure, DNS, and CI/CD while quarantined to my study with a sickness bug.
Exploratory thinking partner versus execution tool - the two distinct ways I've learned to work with AI, and when each mode actually works.
Cardboard muffins, missing babies, and reward hacking - the predictable problems you'll hit when building with AI tools.
From hating chatbots to building production tools with AI - what changed wasn't just the technology, it was understanding when to build jigs.
Seth Godin's Purple Cow isn't about marketing products - it's about getting people to adopt new systems. Finding the early adopters, making things remarkable (or at least easiest), and catching workarounds before they become normal practice.
Why work takes longer than it should - the five thieves that steal our capacity without us realising it.
Who decides what to build? Who says no? Who owns the trade-offs? Learning about product ownership through projects that worked and projects that didn't.
Sometimes you need to plan everything upfront. Sometimes experimentation is dangerous. Sometimes sequential delivery is the only sensible option.
MVP isn't about building a bad version - it's about learning what matters first. From paper forms to real-time alerts in three iterations.
From plane journeys drowning in jargon to building dispatch systems on shop floors - my ongoing attempt to understand what agile actually means.