Writing.
Observations from the front line. Practical thinking on how teams get more from AI without losing what makes them effective. The occasional uncomfortable question.
27 essays · Updated weekly
Filter27 results
- № 012 Jul 2026LeadershipCarbon Removal's Biggest Buyer Is Also Its Biggest RiskAI and tech firms are becoming carbon removal's dominant buyer group, and a market this concentrated in one type of buyer carries a single point of failure.→
- № 0218 Jun 2026PracticeNobody Models the API Bill Until It ArrivesConsumption-based AI pricing has no natural ceiling. Most companies find that out at renewal, after the headcount decision is already made.→
- № 035 Jun 2026ThinkingThe Vanity Metric Anthropic Didn't NeedAnthropic reported an eight times productivity gain and flagged its own metric as misleading in the same paragraph. That's worth sitting with.→
- № 0423 May 2026ThinkingEfficiency Was Supposed to Be the AnswerAI keeps getting more efficient, and the argument that this makes it sustainable gets weaker every time it does.→
- № 0510 May 2026PracticeWhat You Need to Know Before Handing Claude to Your TeamRunning an AI pilot inside a real organisation teaches you things vendor documentation doesn't cover. Here's what actually matters.→
- № 069 May 2026ThinkingAI Is a Multiplier, Not a Headcount CalculatorThe first question most organisations ask about AI adoption is the wrong one. Cutting headcount is not the point. Changing what your team can do is.→
- № 075 May 2026ThinkingThinking Out Loud, or Just Approving?The real risk from using AI isn't that your thinking gets worse. It's that your thinking stops happening while your outputs stay the same.→
- № 0814 Apr 2026CraftWorking smarter with AI: a builder's guide to staying in controlAI tools are powerful. But if you don't understand how they work, you'll burn through context, lose coherence mid-session, and wonder why your builds keep going sideways. Here's how to work with the grain of the tools, not against it.→
- № 0931 Mar 2026ThinkingThe Prototype Is Not the ValidationVibe-coding has removed the natural friction that used to kill weak ideas quietly. Teams now need a different kind of pushback.→
- № 1026 Mar 2026CraftThe Code Quality Question Nobody Is Asking ProperlyAI-generated code forces us to separate the standards that protect outcomes from the ones that protect identity.→
- № 1124 Mar 2026CraftBuilding Fast Is Easy When Nobody's Depending on It YetAgentic coding workflows are compressing timelines in real ways. But most of the posts celebrating this share a variable they consistently leave out.→
- № 1224 Mar 2026ThinkingThe Canvas Is Not a FeatureI switched from ChatGPT to Claude and haven't looked back. Except for one thing, and it's worth being precise about what that thing actually is.→
- № 1322 Mar 2026PracticeLate to Claude Code. Or just in time.Feeling late to a new tool is rarely the right frame. What transfers matters more than what's new.→
- № 1421 Mar 2026PracticeDemocratising AI Is the Easy PartGiving everyone access to AI tools solves the wrong problem if the knowledge those tools need stays locked in people's heads.→
- № 1519 Mar 2026PracticeAI Gives You the Ability to Run Ten Things at Once. That's Not Always a Good Idea.AI expands what you can run in parallel, but cognitive load doesn't scale with capability.→
- № 1617 Mar 2026ThinkingThe Bar MovesThe real risk of AI in the workplace isn't replacement — it's that the bar for what one person delivers quietly shifts upward.→
- № 1714 Mar 2026CraftThe Tooling Didn't Fix the InstinctBetter project management tools haven't solved the core problem: the instinct to build everything before you need it.→
- № 1813 Mar 2026ThinkingStrategy Is Not a GoalConfusing strategy with goals is mostly harmless in a team meeting. In a boardroom, it quietly costs you trust.→
- № 197 Mar 2026LeadershipFull Stack LeadershipThe leaders who struggle in smaller, autonomous teams are the ones who can only operate at one layer.→
- № 2028 Feb 2026ThinkingConfident Answers Are Not Green LightsAI answers sound authoritative even when they're wrong. The problem isn't the model — it's outsourcing the validation.→
- № 2128 Feb 2026ThinkingQualifying, Not OutsourcingWhen AI does the work in two hours that used to take months, the unsettling part isn't the speed — it's what the speed reveals about where your value actually sits.→
- № 2220 Feb 2026ThinkingMost AI Debates Compare the Wrong ThingsEveryone's benchmarking model quality. Most organisations should be mapping their data instead.→
- № 2318 Feb 2026ThinkingFive Prototypes, and Still Not LearningAI lets you build five prototypes in the time it used to take to spec one. That's faster building, not faster learning.→
- № 2418 Feb 2026LeadershipWhat Happens When You Apply Product Thinking to ITTaking on the global IT function after 25 years in product has been humbling — and clarifying.→
- № 2518 Feb 2026PracticeWhen 'Not Yet' Points You at the Real WorkThe right response to a strategic blocker isn't always to push harder — sometimes it's to step sideways and build the thing that's already useful.→
- № 2618 Feb 2026LeadershipAutomating Junior Roles Cuts the Leadership PipelineJunior roles have never just been about the task. They're how people learn how decisions actually get made.→
- № 2721 Jan 2026ThinkingAI Won't Replace Product Managers. But It Will Expose Weak Ones.The claim that AI will replace PMs because it can write PRDs misunderstands what product management is actually for.→