Make the cost of AI visible
AI that bills by surprise erodes the trust it needs to be useful. Show the price before the work, meter it as it runs, and let teams set the ceiling.
ReadPrinciples & Practices for Building
The tools of the last decade were built for a world where software had no opinion and AI had no place in the work. We think both of those are wrong.
This is what we believe about building product in the AI era — written as opinionated essays, not documentation. The practices hold up whether you ship weekly or hourly; the principles are the wedge the whole product is built on.
The Method in short
What we believe
Five positions the whole product is built on. Each is an argument, not a feature — and each links to the essay that makes the case in full.
An AI teammate you cannot supervise is a liability. Agents do the expensive thinking — read the backlog, draft the change, write the changelog — and stop at the one irreversible step, where a human says yes. The default is propose-and-wait, never act-and-apologize.
Read the essayAI is the first feature with a marginal cost per use. Hide it and you train teams to fear the thing you want them leaning on. Show the price before the work, meter it against a ceiling the team sets, and read the spend in the same trail as the action.
Read the essayWhen the board and the tracker are different databases, every handoff is re-entry. One schema means a card and an issue are the same row rendered two ways — for engineering, support, design, and ops alike. The handoff becomes a link, not a copy-paste.
Read the essayA stack of three half-tools taxes you on every boundary: re-typing, drift, standups that open by reconciling what the tools disagree about. Consolidation isn't tidiness — it's removing the seams where work and trust leak out.
Read the essaySpeed is not a vanity metric; it's whether the tool gets out of the way. Keyboard-first navigation, instant views, and tight flow let a team stay in the work instead of operating the software that tracks the work.
Read the essayStart here
All essays
AI that bills by surprise erodes the trust it needs to be useful. Show the price before the work, meter it as it runs, and let teams set the ceiling.
ReadA ticket is a request to be processed. An issue is a problem to be understood. The words you use change the work you do.
ReadDeadlines are predictions dressed as commitments. Cycles are a heartbeat. One creates anxiety; the other creates rhythm.
ReadWhen the board and the tracker are different databases, every handoff is re-entry. Share the data and the handoff becomes a link.
ReadAutomation should remove the steps no human should have to remember — not replace the judgment only a human can bring.
ReadTriage is not data entry. It is the moment you decide what your team will and will not care about — so do it on purpose.
ReadA changelog is not a git log with marketing. It is a short note from the team that built something to the people who use it.
ReadStarting work is easy and feels productive. Finishing work is what ships. WIP limits force the difference.
ReadA roadmap that never changes is a lie you agreed to tell. A roadmap that changes constantly without warning is noise. The craft is in between.
ReadBuilding in public is not a marketing tactic bolted on at the end. It is a forcing function that makes the work itself better.
ReadActivity metrics measure motion. Cycle time measures flow. One is easy to game and the other is the thing you actually want.
ReadPut it into practice
The Method is how we think. The product is how you ship it — free for small teams.