AI, software, the structures that are quietly breaking, and ponderings on what could be next. RSS feed ↗
Independent research. Available for advisory, research collaboration, and expert consultation. Get in touch
27 May 2026 · raw .md ↗
AI's costs are industrial. Its revenue mechanisms are dissolving. The collision between these two forces is pushing the most transformative technology in a generation toward regression, through infrastructure dependencies, venture capital, startup absorption, and the capture of open source. Part 1 of a series on the economic forces shaping AI.
9 May 2026 · raw .md ↗
The post-transaction layer is where the deepest capture in agent commerce occurs: not through security, payments, or model training, but through the record a service keeps after the agent has acted. The risk is that the vendor becomes the only party that can explain how the work was done.
21 April 2026 · raw .md ↗
Per-seat software pricing is a twenty-five-year artifact of the human bottleneck. When agents can invoke a capability thousands of times in an hour, the software market sorts into two categories: tool-makers whose usage multiplies, and coordination surfaces whose seat economics hollow out. A sorting test for your vendor list.
14 April 2026 · raw .md ↗
The competition between frontier AI labs is itself the proof that the model layer is commoditising. The labs matching each other in compressed timeframes, the platform owners building model-agnostic infrastructure, and open-weight models closing the gap from below. These three convergence forces reproduce the structural pattern of the browser war, where the winner didn't get to keep the winnings.
25 March 2026 · raw .md ↗
The agent commerce infrastructure is being built right now. Payment rails, trust registries, autonomous wallets. And the capture this time will be about transactions, trust, and the reasoning traces that accumulate when agents do the work
18 March 2026 · raw .md ↗
The AI agent web is still forming. The protocols are open but capture point is authentication and the identity vendors are moving fast
4 March 2026 · raw .md ↗
AI will produce bad code. Companies will realise their mistake. And then they will call the humans back. That prediction is wrong.
