AI Innovation Speaker
The interesting question about AI innovation is no longer what the models can do. Everybody has access to roughly the same capability now, which means capability has stopped being the differentiator, and the question that replaced it is harder: who can get anything into production and keep it running there?
What the talk covers
Scarcity did not slow startups down the way the coverage assumed it would. It changed who wins.
The entrepreneur extinction event is the long version of that argument. Constrained compute pushed builders toward leverage instead of headcount, and a small team that picks its problems carefully now outbuilds organisations it could not have competed with a few years ago. The asteroid in that title is deliberate, because the useful question for a founder is not whether the event is happening. It is which side of it you are standing on, and the honest answer usually depends on choices about tooling, scope, and hiring that you are making this quarter, not on anything structural about the company you happen to have built. Size stopped being protection somewhere in the last two years. What replaced it is how quickly a team can turn a capability everyone has into something that runs, and that is a much less comfortable moat to sit behind.
The same collapse in cost applies to the work that surrounds building. Industry research, competitive analysis, and customer understanding used to be weeks of effort gated behind a budget. Using AI to unravel industry competition and target users walks through compressing that into hours, which changes who is allowed to hold a considered view of a market before committing to it.
Then the talk turns to what did not get easier. Production is still hard, and the pilots that die rarely die because of the model. They die on qualification, on missing patterns, and on quality gates nobody built.
The counter-position worth holding is that startup architecture is a leading indicator. What frontier teams build now is where enterprise budgets land in a year or two, which makes their choices about agentic systems and inference economics readable as a forecast rather than as news.
Who it is for, and what the room leaves with
Founders deciding where to place a small team against much larger competitors, executives deciding what to fund and what to stop funding, and technology leaders who need to tell a durable shift apart from a loud one. They leave able to read startup architecture as an early signal, and with a clearer sense of why the bottleneck inside their own organisation is delivery capacity rather than demand. Other talks are listed on the speaking page.