Rampell: AI easily handles integer math benchmarks like AIME but struggles with proofs
Alex Rampell · Reid Hoffman on AI, Consciousness, and the Future of Labor · Oct 20, 2025 · at 30:16
Alex Rampell, general partner at Andreessen Horowitz, explains why standard math test benchmarks overestimate AI capabilities compared to generating rigorous logical proofs.
0:00 / 0:18exact quote · 18.8s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“If you look at the progression of AI, there is the AIMI, the American Invitational Math Examination, where you, the answers are all just like three, it's just integers. It's like zero to 999 is the answer. And then, of course, you can keep trying different things, and then you either get the right answer or you don't, and it's very, very easy to do that, whereas once you get to proofs, very, very hard.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Alex Rampell
Rampell: Artificial intelligence technology is currently under-hyped
“This is actually why I am convinced that AI is under-hyped.”
Rampell: Prediction markets gather information more efficiently than the CIA
“And this is a much more efficient truth machine than, like, the central intelligence agency, if you think about it.”
Alex Rampell asserts Google is artificially inflating Gemini daily active user metrics
“And I think the metrics that they're kind of faking right now is it's like what percentage of our users are using AI and that's not using Gemini as a DAU. I'm sure of that. It's just like they have like an AI summary, something, something in there, but they're…”
Rampell: Traditional banks will never be able to build software
“Well, and that's the thing, it's like, you know, software, like I'm very confident that all of the banks Cannot build software, will not be able to build software, will not build software.”
Rampell: Zillow's lead gen model is terrible and deserves -100 NPS
“And what Zillow and others do is they just do lead generation. It's a terrible business model from a consumer perspective. Like it should have like negative a hundred NPS because they're just, you're just getting called all day by real estate agents.”
Rampell: Coupon extensions like Honey steal affiliate attribution at checkout
“And then it's like, do you want a coupon code? Oh, why? Yes, I do. Why would I not want a coupon code? 10% off, click here. You go click here. What does it do? It redirects you to an affiliate page. It puts a cookie on your machine. It redirects you back to th…”