Writing
Blog
Thinking about AI product leadership, building with AI tools, governance, and the economics of shipping AI in production.
Writing
Thinking about AI product leadership, building with AI tools, governance, and the economics of shipping AI in production.
Showing 61–72 of 77 articles

AI commoditises execution. The scarce resource is knowing what to build, for whom, and when to stop. That's taste, and it's the career bet worth making.

Most teams evaluate agents with manual chats and gut feel. A practical framework for eval suites that let you ship, starting with 20 examples, not 20,000.

Agentic coding compressed the PM translation layer to zero. The three skills that matter now: problem shaping, context curation, and taste as judgment.

AI strategies fail because leaders set direction for a capability they have never used. You cannot strategise well for a material you haven't touched.

AI does not replace jobs. It replaces tasks. That distinction changes everything about how you plan your career, your hiring strategy, and your org chart.

If your AI roadmap succeeds, customers need fewer seats and you earn less revenue. The fix: price around the units of work completed, not user logins.

A manager model checking every worker output increases unit cost by 2,500%. The fix: a spot-check architecture that can save 75% of your token margin.

The shift is from prompt engineering to designing multi-agent hierarchies: AI managers overseeing AI workers that operate invisibly in the background.

AI is not a feature, it is a new compute paradigm. Bolting GenAI onto legacy platforms destroys unit economics. If the AI is optional, it's a gimmick.

Google's A2UI signals the end of the chatbot text wall. Agents that render native UI components instead of paragraphs change what product teams build.

A 5-step agent at 95% accuracy per step is only 77% reliable. The path forward isn't better agents, it's narrower ones. Three rules for workflows that ship.

Google is testing native property listings in search. AI broke the constraint that stopped them in 2011. The aggregator model dies when the user is an agent.