Ideas for the people who build what commerce runs on

I edit The Rigorous Builder, a journal of long-form essays on AI, product and commerce systems. Every essay has to pass one test: it should be useful to a graduate in their first month and to a chief executive in the same sentence, and it should leave each of them with one distinction they can take into a meeting.

The Real Challenge of AI Agents Is the Right to Act. An approval records a decision about a world that may already have changed. Agent authority as a bounded, expiring grant.

AI Product Readiness Is a Rate Claim. A demo proves an AI product can work once. Launch needs a defended bound on how often it fails, on which work, and who catches it.

Designing AI Interfaces for Decisions That Matter. AI experiences built around intention, informed decisions and recovery when work moves across tools, people and time.

The Economics of AI Begin After the Answer. AI investment judged by completed work, bottlenecks, review capacity and the cost of changing course.

Choosing an AI Model Is a Decision About the Whole System. Workload, uncertainty, cost and reversibility, with worked examples that show why rankings change.

GPT-6 Astra Explained. The same model scored 62.7% and 98.6% on the same test. The five controls that decide which number you get.

Vibe Coding, Explained: The Code Was Never the Hard Part. AI can now build the screens, the logic and the API in an afternoon. The failures live in the layers a demo never shows.

Leading AI-Enabled Teams. AI multiplies how fast a team can produce work, not how fast it can accept it. Responsibility, review and coordination, redesigned.

AI Skills That Last. How individuals and organisations build judgement, verify AI-assisted work and grow expertise that outlives the tools.

The Future of Product Management. A focused AI toolkit, useful applications, won customers and lasting value.

The Future of Software Engineering. AI made code cheap to generate and left the cost of trusting it where it was.

The Product Manager's Guide to Reading a P&L. Every feature moves one line of the P&L and quietly leaks into another. How to name both, with the numbers.

The Ten Stages of AI Expertise, Measured by the Errors You Can Catch. Tool fluency plateaus within months. Judgement compounds. A ten-rung ladder built on the errors you can catch and own.

Why the Second Marketplace Costs More Than the First. Connecting a marketplace gets cheaper each time. Reconciling it gets more expensive, and most business cases leave that out.

AI in E-commerce: Building Customer Trust. Shoppers judge an AI assistant on what arrives, not on what it said. The seven moments where trust is kept or broken.

What Must Be True When an AI Agent Clicks Buy. The eight bindings of agentic commerce, from product identity and inventory to payment, fulfilment, returns and loyalty.

Why Commerce Migrations Fail After Go-Live. Cutover moves the data. Stabilisation proves the business still runs. Plan, fund and staff them as two projects.

I research and draft every essay with AI assistance. The argument, the judgement and the final edit are mine, and each essay says so.