What About Replacing Y Combinator Partners with LLMs?
Y Combinator has posted their fall Request for Startups, and one of the RFSs caught my eye: Using LLMs Instead of Government Consulting. The basic idea is that it’s feasible to replace these consultants with LLMs; the key thesis is “LLMs today are so good that they can already do the jobs of many consulting firms.”
That got me thinking — given how good LLMs are, could we replace partners at places like Y Combinator with LLMs?
So I (well actually ChatGPT) wrote an RFS:
Using LLMs Instead of General Partners at Y Combinator
RFS by Steven Strauss, Ph.D. / ChatGPT — August 2025
The venture capital industry stewards about $1.2 trillion in U.S. assets and pays ≈2% in annual fees — roughly $24 billion every year — to partners and platform teams, even though half of all general partners are, by definition, below median in performance.
Those same workflows — thesis building, sourcing, diligence, and founder support — are largely text- and conversation-based, precisely the arenas where large language models now excel.
Recent “Wade Test” experiments show employees can’t reliably tell a fine-tuned LLM from a real C-suite leader. That means we can train an LLM on YC partner communications — office hours transcripts, Demo Day feedback, investment memos — and create a synthetic general partner that:
- Reads decks
- Drafts investment committee memos
- Flags portfolio red flags
- Facilitates introductions
- Even delivers empathetic founder therapy in seconds, not weeks
Unlike humans, the model can be explicitly programmed to stay within ethical guardrails :-)
This could be a huge opportunity for founders to build these LLM-native general partners and return billions in saved fees to startups and LPs.
Important Notice: This blurb was prepared by Steven Strauss and was not prepared or approved by anyone at Y Combinator.
Additional Context
BTW, this is not perhaps as flippant as it sounds — a couple of groups are experimenting with AI-driven investing, see for example No Cap.
For those of you who want more information on this, I had ChatGPT do deep research on it, and used that to have NotebookLM generate a video. You can access the NotebookLM with the video at this link and the video is shown above.
(For what it’s worth, my bet is many VCs/Angels are doing this on an ad hoc basis, but it does seem like a real opportunity)
Comments are welcome!
