Notes that held up.
I think in public, mostly about products, AI and the strange new economics of trust. These pieces earned their keep. Each links to the original, unedited.
Perplexity’s real product is curation
Everyone reads it as an AI model marketplace. Look closer and the product is editorial judgement: what to surface, what to trust, what to skip. The models are staff. The taste is the company.
Read on LinkedIn ↗ October 2025When AI fails, it is trust that crashes
The Cruise robotaxi case, read properly. The failure was never just the algorithm; it was everything around it. A cautionary tale for anyone shipping AI into consequential settings.
Read on LinkedIn ↗ December 2025The ROI of AI agents
Agents are not a cost-cutting story. They change the physics of decision-making: how many options you can evaluate, how often, and at what depth. That is where the returns hide.
Read on LinkedIn ↗ June 2025The hidden crisis in team performance
Gallup’s 2024 numbers say the quiet part loudly: the problem is not effort, it is design. What organisation builders should take from the data, before culture becomes a slogan.
Read on LinkedIn ↗ April 2026AI as a blind-spot finder
I put an AI growth assistant on my own job search, scanning listings daily for what I systematically overlook. Using AI on yourself is the cheapest usability study you will ever run.
Read on LinkedIn ↗ June 2023Crossing the chasm, applied
Geoffrey Moore’s early adopters and early majority want different truths from the same product. Notes from selling disruptive products to people who did not ask for disruption.
Read on LinkedIn ↗Proof I use the tools.
A product person who does not build with AI daily is reviewing a restaurant from the car park.
A personal AI copilot on Microsoft Azure, summoned from WhatsApp. It knows my work, drafts in my voice, and taught me more about agent reliability than any article could.
104 million tokens through Cursor, and usage in ChatGPT’s top 0.1 per cent. Not badges, just evidence of where my hours actually go.