Tom Blomfield
person concept tool org talk claim — click a node to jump to its page; hover an arrow for the relation
Tom Blomfield is a technology entrepreneur and investor (associated with Y Combinator) who advocates for reconceiving companies as AI-driven, self-improving systems rather than using AI merely as a productivity tool for individual workers. His framework, partly inspired by a talk from Diana and ideas circulated by Jack Dorsey, argues for a fundamental restructuring of organizational design around AI loops.
Critique of the Copilot Model
Blomfield's central provocation is a rejection of the prevailing "copilot" paradigm. He argues that "the copilot mental model — using AI to make individual workers incrementally more productive — is a broken way of thinking about AI", calling it a fundamentally limited frame. In his view, treating AI as a tool that augments individual humans misses the deeper structural transformation available to organizations.
The Self-Improving Company
Instead, Blomfield advocates for the Self-improving company model: "you can reimagine what a company is as a set of recursive self-improving AI loops" that improve without human intervention. His canonical example is a monitoring agent that autonomously identifies failed queries, generates fixes, and deploys them overnight — which he characterizes as a genuinely self-improving system rather than just a productivity tool.
A prerequisite for this model is making all organizational activity legible to AI. Blomfield argues that recording all organizational activity is the foundational prerequisite for building a self-improving company.
Domain Knowledge and the Company Brain
Blomfield places high value on Domain knowledge extraction — "extracting the domain knowledge from your company and defining it as context or a set of skills". He frames this accumulated knowledge as the Company brain: "all of your data, all of your emails, your DMs, the skills, the know-how — that is like the company brain", treating it as the durable, appreciating asset of the organization.
Complementing this is his advocacy for Software ephemerality: "treat the software as ephemeral. You can generate it. You can regenerate it." In his framing, business context and skills are the valuable, lasting assets of a company while software built on top of them is ephemeral and should be discarded and regenerated.
Organizational Structure and Headcount
Blomfield draws sharp conclusions about org design. On the resource question, he advocates for Token usage over hiring: "burn tokens, not headcount", and predicts companies will soon be constrained by token usage rather than employee count. He cites empirical support from his YC vantage point: YC companies are reaching demo day with approximately 5x more revenue per employee than 18 months ago due to AI.
On roles, he is direct: "I think middle management is done", arguing that AI should handle the coordination function that Middle management previously served. The corollary is that Individual contributor roles become the norm: "everyone just has to be an IC now, a builder, an operator".
The Role of Humans
Despite the AI-centric framing, Blomfield does not eliminate humans entirely. He argues that humans in an AI-powered company serve primarily at the edges where intelligence makes contact with reality, handling novel situations, ethical considerations, and high-stakes moments. This positions humans as boundary handlers rather than coordinators or routine executors.
Points of Disagreement
No speakers in the available material directly contest Blomfield's claims. His positions — particularly that middle management is obsolete and that token spend should replace headcount — are stated as confident predictions rather than explored debates within this talk. The strongest implicit tension is between his vision of largely autonomous self-improving loops and the retained human role at "the edges," though he does not treat this as a contradiction.