Writing
How we think about leading engineering teams now that AI writes so much of the code.
Leading engineering teams
What to bring when your board asks if AI is paying off
The question is coming, and usage numbers won’t answer it. The three questions that will, in the order that matters, and what you can honestly walk into the board meeting with.
Before your AI acts on a number, check it against your own
What it takes to trust the numbers your AI acts on, from agreeing what each one means to checking it against your own systems before anything goes live.
What has to happen around an AI agent before it pays off
If you’re weighing a tool that builds AI agents for your team, three things around the agents decide whether they pay off, and each one needs someone who owns it.
When AI makes a few engineers faster, spreading it to the whole team is a leadership job
Why the gains from AI stay with a few engineers, what the research says spreads them, and the weekly habit that turns one engineer’s good practice into the whole team’s.
An engineer running AI agents is managing a team, starting with themselves
What running several AI agents at once does to an engineer’s grasp of their own work, how that drift turns into a team that no longer understands its system, and what leaders can change.
Why slowing down first makes engineering faster
Why the first weeks of our work with a CEO slow things down on purpose, what that means for an engineering team, and why it matters more now that AI writes so much of the code.
How we build
The methods and guardrails we use when we build with AI.
How AI-written code gets proven before it counts
The standard of proof we run on our own builds, from challenging a risky plan before it runs to opening the finished page in a real browser, and the smaller version your team can start with.
Does your plan match your code?
How every statement in a plan, a whitepaper and a spec gets checked against the line of code that decides it, so you know what’s built, what isn’t and where the documents disagree.
Reporting engineering progress to your board, with evidence behind every line
How to write an engineering update that keeps your board’s confidence while every claim holds up under questioning, and how to keep what’s built apart from what still needs time or money.
Testing a product bet before your customers do
How a simulated population of 3,000 households, built from published research with every assumption labeled, showed where Debt Hacker’s approach breaks and who it fails, before customers carry the risk.
What happens when an email tells your AI to do something
Why every coding agent that reads a ticket or a pull request needs the rules two of my own assistants follow: assume anything they read may carry a planted instruction, and keep the tools that could do damage locked.
How an AI agent tests the pricing at Lost Horizon
How an AI agent tests the pricing at Lost Horizon, the glamping retreat I designed, checking each change against a year of data and the market, with the calls that matter still coming to me.