All Gas No Brakes

Notes on building — a series in 28 parts, 28 out so far.

  • Compounding goes both ways

    XXVIII — Trust, culture and technical debt all accrue quietly. Only one of them is usually being tracked.

  • The founder is the thermostat

    XXVII — Your state sets the ambient temperature of the company, which makes recovery an operational concern.

  • Craft compounds

    XXVI — Small details add up to trust. So does their absence, on the same schedule and in the same direction.

  • Scale the right parts

    XXV — Scaling isn't more of the same. It's finding out which constraint breaks first and rebuilding only that.

  • Features need stories

    XXIV — A feature list tells people what exists. A story tells them what changes. Only one of those gets acted on.

  • Small, honest experiments

    XXIII — State the hypothesis, the metric and the smallest test that could move it. Most teams write only the third.

  • Failure is data

    XXII — Cheap failures are worth having. The point is the recovery system, not the slogan.

  • Hire for judgment

    XXI — Early hires either multiply the number of good decisions being made, or they add throughput. Only one scales.

  • Attention is the budget

    XX — Every feature spends something finite. Most roadmaps account for engineering time and not for the user's.

  • Empathy is discovery, not projection

    XIX — Ask what they were trying to do, not what they want you to build. The difference is most of research.

  • Metrics that change behaviour

    XVIII — Most dashboards report on decisions already made. A useful metric changes what you do this week.

  • Stewardship in AI

    XVII — Safety constraints cost you something now and buy the only thing that makes the category viable later.

  • Decisions are experiments

    XVI — The choices that matter get made without the data. Optimise for what you can undo.

  • Design as philosophy

    XV — Design decides what people can do and, eventually, what they think of as possible. That's a bigger claim than it looks.

  • Process is a tool

    XIV — Frameworks are scaffolding. The failure isn't picking the wrong one — it's forgetting it was scaffolding.

  • Rigour and resonance

    XIII — Research gives you the map. Craft gives you the reason anyone follows it. Neither substitutes.

  • Systems are not their parts

    XII — Optimising every component individually is a reliable way to make the whole thing worse.

  • Usability is an ethical position

    XI — A confusing interface spends someone else's time and attention. That's a transfer, and it has a direction.

  • Designing for mood

    X — The same screen lands differently depending on what happened before it. Most of that isn't yours to control.

  • Users invent the product

    IX — The uses you didn't plan for are where the market actually is — and most teams treat them as bugs.

  • The invisible tool

    VIII — A good tool disappears into the intention. Getting there costs more than adding features.

  • Value before polish

    VII — Heckel's law: when perceived value is high, people forgive the interface. It's a licence with limits.

  • Designing for foragers

    VI — People hunt information the way they hunt anything else: follow the scent, abandon the trail when it weakens.

  • Authority without permission

    V — Distributed roles work when the process is real. Otherwise you've replaced a bottleneck with a vocabulary.

  • Products are relationships

    IV — Adoption, habit, personalisation, plateau, abandonment. Every product is somewhere on that curve.

  • Everything starts as a fragment

    III — Projects don't begin whole and they don't end. The friction is information, not punishment.

  • Intentional incompetence

    II — Competence is comfortable and comfort is where plasticity dies. Keep choosing things you're bad at.

  • Luck is biased

    I — Fortune isn't random. It's weighted toward people already moving — a smaller claim than it sounds.