AI•Sep 19, 2025•English
VCBench: Are LLMs better at VC investing than humans? An Oxford student benchmarked LLMs from the perspective of predicting founder success and argues they outperform people.
If everyone eventually runs LLMs to decide whether to invest, then where does a VC firm's edge come from?
- Having your own investment criteria that let you invest where others look away. In Moneyball terms, when everyone else looks at batting average, I look at on-base percentage.
- Picking a good company and actually getting allocation into that company are completely different skills. If you cannot do the latter, great investment evaluation does not matter.
- Persuading founders to take money from you, and engineering the relationship and trust that makes that possible, is still human work.
The experimental design in this paper is overly simplified, so the limits on generalization are obvious. Still, it raises the right question. AI and LLMs force every profession to reexamine its essence.
