The Ledger
Every statement that passed quotation and attribution checks. Mix any filter with any other: certainty 1/5, debate potential 5/5, or both at once.
why aren't all 7 resolved? a statement only gets an assessment when the public record can support or contradict it. opinions and what-ifs never can, and 0 checkable ones are still open, waiting for their date. predictions held up or didn't; assertions are supported or contradicted. on every card: ▮▮▮▮▮ certainty · ▮▮▮▮▮ debate potential. speakers are clickable
Pope: Groq and Cerebras are uncompetitive on dollars per token
“And then there's the Grok and Cerebris that are much better at latency because they've got this the SRAM, weights are in SRAM very low latency. The problem is, and the challenge when you go to a Grok or a Cerebris system is that the throughput you get there, i…”
Sundheim: Nvidia is not expensive and public AI stocks aren't in a bubble
“I would tell you, like, NVIDIA is, we can have a debate about what the earnings are going to be in a few years but it's not expensive.
Like, NVIDIA trades at twenty-something times multiple.
that's, you know, I think within reason, and I don't think there's…”
Collison: Silicon Valley over-valorizes unrelated second acts compared to vertical focus
“I think second acts are kind of a bit over-valorized in Silicon Valley somehow, where everyone talks about, oh, we want to do, like, a new AWS is kind of the coolest thing ever, where you do a second thing that is completely unrelated to the first thing, versu…”
Ricks: Eli Lilly is building pharma's largest biological supercomputer
“Building a supercomputer on our, on-prem for us, really just to run proprietary drug discovery models. We think it's the biggest biologically focused supercomputer there is. And certainly the biggest pharma's done. Yes. With B 300 latest chipset.”
Pope: 4-bit precision will probably be the primary AI format, like Nvidia
“We think probably the main thing will be similar to where NVIDIA is at, which is four bit precision.”
Pope: Google's decision to build TPUs specifically for neural nets paid off
“They at least had the option, like the opportunity to design the TPUs for neural nets at least rather than graphics applications like NVIDIA. And so the overall architecture starting with single core doing what was at the time reasonably large systolic arrays …”