Keep pulling the thread on Jack Dorsey.
AI is fundamentally breaking the traditional hierarchical organizational structure of companies, a model inspired by Jack Dorsey's thinking.
Companies can be reimagined as a set of recursive, self-improving AI loops rather than as traditional hierarchical organizations.
Y Combinator's internal AI system can autonomously identify failed queries, write code to fix the issue, submit a merge request, have an AI agent review it, and deploy the fix overnight.
The new paradigm for building companies is to prioritize spending on AI model tokens over increasing employee headcount.
Y Combinator is observing that companies reaching its Demo Day now generate approximately 5 times more revenue per employee compared to 18 months ago.
The speaker predicts that companies will soon face constraints based on their AI token usage rather than their employee headcount.
The speaker believes the role of middle management is becoming obsolete as AI can handle the coordination tasks they traditionally performed.
To build a self-improving AI-native company, it is essential to make the entire organization "legible to AI" by recording all communications and activities.
Using 2,000 hours of recorded office hours, a Y Combinator partner named Harsh generated a new, 150-page YC user manual over a single weekend that is considered dramatically better than the previous version.
The primary value in a modern company lies in its accumulated business context and skills, not in the software itself, which is becoming a disposable commodity.
A common view one year ago was that AI's primary utility was productivity enhancement, such as Copilots making engineers 20% more productive.
The speaker believes that viewing AI solely as a productivity tool to enhance old ways of working is a flawed mental model.