Founder · Investor · Writer
Ideas on intelligence, ownership, and the systems quietly rearranging the world.
I build and back companies across software, commerce, and applied intelligence. This journal is where I write through the shifts before they become consensus.
We are leaving the era of AI as a feature. The winning products will reorganise themselves around judgement, not generation.
When competent code becomes abundant, deciding what deserves to exist becomes the scarce capability.
Autonomy will not scale on demos. It needs identity, permissions, memory, receipts and reliable ways to fail.
Average work gets cheaper first. The edges—trusted commodity and exceptional craft—become more valuable.
Code, media, capital and machine labour are collapsing into one operating system for small teams.
Software is moving from passive canvas to active participant. That changes the contract between user and machine.
The cost curve for cognitive output is bending. Business models built on hours will feel it before balance sheets do.
AI makes industry-specific context more valuable, not less. The deepest workflows are becoming buildable again.
The algorithmic feed is saturated. Trust is moving back towards people, private graphs and useful artefacts.
The shape of a serious company is changing before the language around company-building catches up.
The technology becomes transformative when it disappears into a dependable workflow.
Energy, chips, networks and sovereignty are turning data centres into geopolitical infrastructure.
Software does not become free. The expensive parts move to specification, integration, trust and maintenance.
If anyone can build the interface, the company must own a deeper reason to continue existing.
Intelligent products must continuously negotiate intent, authority and uncertainty with their users.
AI investment is colliding with infrastructure timelines, regulation and organisational memory.
The best early users do not need a vision explained. They need the missing instrument built properly.
Model quality attracts attention. Cost structure, workflow position and value capture decide the company.
The frontier model race is concentrated. The application layer remains open—but not indefinitely.
Conviction is not stubbornness. It is a testable model of why a product should exist and what must be true.