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 108 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
Dettmers: AI hardware has maxed out and won't get faster
“The hardware is maxed out. We have no new technology. We can make it easier to manufacture and a little bit cheaper, but not faster. And we have maxed out on the additional features.”
Amr Awadallah claims Vectara has solved the LLM hallucination problem
“I mean, you do have a human in the loop, so you can still qualify it, but we want it to be minimized to zero, and that's the problem that we solved.”
Feldman: Cerebras sales were 10x higher than Groq's at acquisition
“And we were the fastest at it, and the largest, and, you know, our sales were more than 10 times the Grox, and they paid twenty billion dollars for the number two collector.”
Zeghidour: Kyutai's Moshi remains the only full-duplex conversational AI model
“Moshi, that is still to the day the only full duplex model.”
Hebbia was the first company to productionize RAG in 2020
“Hebbia were actually the first to turn that into a product. So it's like a very close thing to my heart. So back in 2020, we were the first people to actually productionize it, roll it out.”
Replit claims to be only company offering full-stack AI app generation
“We're like, we're the only company that that's done that, which is like the full stack experience from the code to the databases, to the cloud deployment.”
Duderstadt: No top benchmark AI embedding models release their datasets
“And as far as we're aware, none of the other, you know, top models on any of these benchmarks have gone out and released their data sets.”
Zhou: Lamini is the only platform running LLMs on AMD GPUs
“We are the only folks who can actually run your language models on top of AMD AMD GPUs.”
Lamini's AMD support unlocks 20,000 GPUs, enough to train GPT-4
“So that unlocks, what that means is that unlocks about 20,000 GPUs readily available today for enterprises to be able to use, and to get a sense of what that means you can train GBD-IV.”
Elsamadisi: Redshift, Snowflake, and BigQuery perform identically on long tables
“The speed improvement on Redshift and versus Snowflake versus BigQuery becomes interchangeable. Because what they handle is for those edge cases, we have like a bunch of, like Snowflake is really good at a bunch of small tables everywhere trying to load a memo…”
A trained CTRL-Labs band will not work at all on another person
“This is a very personal model that is custom to just you. When you put on the band after you've trained it, if you put it on somebody else, it doesn't work at all. It is, and it's not like design. That's just the way neural interfaces are required to work. The…”
Uretsky: Kelsey Hightower rated DigitalOcean's Kubernetes UX superior to Google's
“Kelsey Hightower, one of the core contributors to Kubernetes and over at Google, our friendly company, actually surveyed a bunch of different providers delivering Kubernetes services and said that DigitalOcean provides the best developer experience And is the …”
Bordes: 2016 image recognition AI uses Yann LeCun's 1993 neural architecture
“This is exactly the same architecture of what's being used right now.”
Amazon Web Services holds about 90% market share
“You have a company that's gonna, ah, it's on its path to be, and it is, about a 90% market share player.”
Gutman: 74% of nighttime ER visits are unnecessary
“74% of people that go to ERs at night don't need to be there.”
eHarmony couples had nearly half the divorce rate of offline couples
“The couples that met on eHarmony also had a low divorce rate, so nearly half of what the marriages that met actually offline.”
Socher: Parallel Bio secured FDA approval to skip certain animal trials
“That company alone has already gotten FDA approval to skip certain animal trials.”
Prince: More money goes to music creation per GDP today than ever in history
“We can quibble about are the right people getting the money or not, but there's more money going into music creation today on, on, on per capita per GDP. Than any time in human history by far.”
Patel: The US has not built wholesale new power capacity in 50 years
“America's not built power for 50 years really, right? It's like converted from coal to gas and like things like this, but like really just have not built wholesale new power on a large scale.”
Patel: Over 99% of OpenAI's spend is likely compute
“99 plus percent of their spend at the company is probably just compute.”
Douglas: Anthropic models autonomously replicated the Claude.ai website in hours
“And in this case, the model replicated Claude.ai with artifacts, with everything else I can't quite remember how long that one took. Maybe a couple hours to do.”
Kiela: FAISS was the first vector database
“In the initial paper, we used a vector database or a face. So the words vector database didn't exist at the time. But so face was the first vector database.”
Replit Agent was the first AI programming agent on the market
“So we got to September, we launched it in early access, and at the time, there was no other programming agent on the market, so it was really the first.”
Vector RAG requires full re-indexing for single-file updates, Habib claims
“And when that policy, even one policy gets updated, you're throwing away the entire embedding store and starting over.”
Socher claims Coca-Cola is the world's largest milk producer
“Coca-Cola, which is actually the largest milk producer in the world, a lot of people don't know that, but they do a lot of things”
Glyman: Ramp is the only platform indicating whether companies overpay vendors
“Today ramp is the only place where you cannot just spend money on a vendor, but we can tell you if we're, if you're paying too much.”
Emi Gal: No existing MRI datasets combine healthy and diseased longitudinal data
“There are no data sets out there with a blend of healthy people and diseased people with longitudinal measures over time”
Socher: You.com launched the first internet-connected LLM search chatbot
“Then we were the first last year to bring chats into the search context also by having the first LN shipped to the world, as far as we know that connected and had a connection to the internet.”
Moses: Monte Carlo uniquely offers out-of-the-box cross-system observability
“So we're actually the only sort of product and market that you can connect today to those sort of different systems and sort of automatically out of the box, get an overview of what the health of your data looks like and sort of observability for your data on …”
Pesenti: Facebook employs 30,000 content moderators
“So we have actually 30,000 moderators that try to really understand if the content that's published on a platform satisfy our policy, but 30,000, given the number, the amount of content, and you're talking about billions of pieces of content every day is not e…”
Cerebras built the first supercomputer designed end-to-end for AI
“And it's also the first supercomputer that has been designed from beginning to end for artificial intelligence and deep learning.”
IT services growth outpaces product growth because enterprise products suck
“There's a reason why that, like, the growth in services is going higher than the growth in, in vendor products, man, because, like, products suck, and they're hard to use, and they change all the time, and like, have somebody else, you know, arbitrate that for…”
Endor outperformed nearly 1,000 teams on a Kaggle benchmark
“Almost a thousand teams participated. And they were given three months, and you can see that the team who won the first place after us, they submitted almost 200 models, so they actually worked pretty hard, and as you can see here, our accuracy is significantl…”
Muglia: Snowflake delivers better price performance than Amazon Redshift
“We were even actually found ourselves to be more price performance Than other cloud solutions like Redshift.”
Liu: Airtable built the first real-time synchronized relational database engine
“In Airtable, we spent a lot of time up front investing in the first real-time synchronized relational database engine out there.”
Gutierrez: Credit Card Act slashed addressable credit card market to 17%
“So the Credit Card Act by itself, before, ah, about 50% of households in America were part of the addressable market for credit card lending, and after the Credit Card Act, it's more around 17%.”
Eric Frenkiel criticizes MongoDB for relying on a global write lock
“Mongo also has a global write lock. So one writer will block everyone else until it completes.”
Japan's sensor network stopped every bullet train 43 seconds before 2011 quake
“Every bullet chain And every factory in Japan stopped 43 seconds before the earthquake hit.”
Evans: Corporate audit costs stayed flat since early 2000s despite software progress
“Average audit data, all sorts of data for audit costs since the Barnes Oakley, which has basically been flat since about since the early 2000, despite everything that's happened in software”
The open-source Lean dataset contains only tens of millions of tokens
“It's only two-digit million number of tokens out there in the open, open world.”
Dettmers: No open-source system supports complex inference across eight-plus GPUs
“There's no open source system that can do that at the moment.”
Palmer: Sutter Hill Ventures incubates all its portfolio companies in-house
“And where Sutter Hill's a little bit unique is that Sutter Hill doesn't invest in pre-existing companies. All the companies are incubated directly out of Sutter Hill.”
Fireworks AI was first to enable function calling for DeepSeek models
“We have been working on function for calling for a long time, and we are the first one to enable function calling for deep seek models.”
Kiela: FAIR was the first team to build a generative RAG model
“Why RAG became the way you name these things is because it's generative, right? So we were the first ones to have a generative model there.”
Misra: Existing text-to-video AI models produce only silent B-roll
“If you look at all the video generation models today that are doing text to video, they're all silent videos, right? There's never anybody talking. It's all B-roll, right?”
Replit built Ghostwriter, the first GitHub Copilot alternative
“So we built the first sort of co-pilot alternative, essentially.”