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 585 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
Tan: YC AI agent conversations are globally viewable by full-time staff
“By default, the agent conversation is actually globally view viewable by any full-time employee at YC.”
Garry Tan runs 10 to 15 parallel Claude Code sessions simultaneously
“I run 10 to 15 parallel Claude code sessions all at the same time.”
Garry Tan: Multi-session AI workflow enables up to 50 PRs processed daily
“When it's ready to land, it lands, and I can do 10, 15, 20, sometimes 50 PRs in any given day, depending on the number of meetings I have in that day.”
Vuong: Physical Intelligence aims to build a universal robotics model
“Our mission is to build a model that can control any robot, to do any task that is physically capable of, and to do so as such a high level of performance that's going to be useful to people in all walks of life.”
Bilgen: Synthetic DNA removes PCR amplification noise from sequencing data
“So what we have done is to add a synthetic DNA into the patient sample that we get before any amplification happens. These synthetic DNA Allow us to know how much amplification happened at different genomic locations. You know, what are the errors that are bei…”
Bilgen: Small R&D teams own products and report directly to co-founders
“We actually have very small you know, research teams. It is essentially principal investigators, like a scientist who is interdisciplinary, who has a small team of, you know, two or three research associates, and they all directly report to David and me, and t…”
Mellata: Variance agents can scrape legacy human review UIs via browser
“So at the moment, the way to integrate with variants can be reverse ETL, API, but very recently we've been awarded this third way, which is spinning up a browser, opening up a really old review tool that was built for a human, pulling that data, and then reaso…”
Mukund Jha: Every Emergent employee must do weekly customer support
“Like one of the things that we do really religiously is everybody talks to a customer once a week, twice a week. Like everyone in the entire company. Everyone in the company. Right. They talk to a customer, everybody does customer support.”
Emergent augments foundation models with fine-tuned verification layers instead of building models
“We don't want to, like, build a Opus 4.5 alternative right away, but we do want to augment it through our custom fine tune verification layers.”
Hodak: Science is submitting its retinal implant for market approval this year
“So we are submitting for approval now. It's not approved on, on the market yet. Hope to have that later this year.”
French-Owen: OpenAI used reinforcement learning on o3 checkpoints for coding tasks
“Well, basically we kind of had like a checkpoint. For I think it was O three, like one of the reasoning models. And then we did a bunch of fine tuning on it and reinforcement learning where it's like, oh, you're given a bunch of questions to like solve these c…”
Garry Tan: Reaching 100% Test Coverage Supercharged Coding Velocity with Claude Code
“I was operating for, like, the first two or three days of my nine days in the wilderness. Like no tests or very few tests, and then one day I was like, all right, today's refactor day. I'm gonna do, get to a hundred percent test coverage, and then I just sped …”
Lapsa: Stoke Space is building fully reusable rockets with aircraft-like turnaround
“Stoke is building fully and rapidly reusable rockets with aircraft-like reusability.”
Lapsa: Stoke Space has raised about $990M to date
“We've raised about nine hundred ninety million dollars to date.”
Lapsa: Stoke Space Aims to Make Every Rocket Part In-House
“We want the ability to make every part on the rocket ourselves.”
Lapsa: Stoke Space built proprietary software Boltline to run factory operations
“So we made the decision very early on. We're going to build our own tool to do all this. We're going to base our operations around it and call it bolt line. And it's been a big part of our success”
Tan: Still a long-term investor in Nvidia despite increasing competition
“You know, I think I'm still a long-term buy and hold, honestly, but for the moment, people are like, oh, well, Gemini's so good, and all the, you know, nobody seems to be NVIDIA only now, and everyone's buying AMD, and everyone's, you know, and TPUs are workin…”
Tan: ARC Prize joins Y Combinator's Winter 2026 batch
“The ARK AGI prize is you know, gonna come in and do the winter 26 batch as a non-profit.”
PostHog is expanding product breadth rather than moving upmarket
“So instead of kind of going up market, we're going just across more of kind of the customer data that people are generating.”
Heller: Casetext pivoted with $20M revenue and 100 employees for GPT-4
“We were like twenty million dollars in revenue. We were doing great. I had like a hundred people and we stopped everything that we were doing and said, we're going to build something totally new based on this new technology.”
Anthropic reverse-engineered cloud provider GPU physical room layouts to fix latency
“We're using a cloud provider, but I think it's kind of, it's not actually that different, because one of the things that was surprising to me is you actually have to understand the literal layout. Like, I remember at 1.1 of my co-workers running a clustering a…”
Nick Joseph: Anthropic hired theoretical physicists who learned programming on the job
“Early on, we hired a lot of people from just, like, all sorts of backgrounds, and I think that people who are just smart and work really hard can learn this pretty fast, but you have to, like, want to. We heard a lot of physicists, for instance. Like theoretic…”
Masad: Replit combines design, engineering, and PM roles into single generalists
“We're building, for the first time, we're building a, like a actual product team, product management team, and our product team is actually made of designers, engineers, and product managers, all, almost always in the same person. So we're trying to merge a lo…”
Legora raises $80M Series B led by Iconiq and General Catalyst
“We raised eighty million dollars led by Iconic and General Catalyst, and, you know, grateful for YC's continued participation, as well as Benchmark and Redpoint.”
Legora operates with intense hours outside of traditional Swedish workplace culture
“We're also very upfront with candidates that this is not a nine to five and we're not the traditional Swedish working environment. We have the good stuff. We have the Fika, but we have. A lot more hunger and, you know, frankly, a lot higher expectations.”
Field: Figma restricts internal product roadmaps to one-to-three-month cadences
“When a team at Figma comes to me with a epic roadmap that they think is perfection, the first question I always ask is, how do we slim that down? How do we make it more bite-sized and test this earlier with our users? So it's absolutely the case that I try to …”
Arne Englander: Model ML aims to be a 10-person unicorn
“We want to be the, maybe not the first, but you know, one of the first, you know, 10 person, billion dollar company.”
Finn: Physical Intelligence uses PaliGemma 3B with flow matching diffusion head
“We took an open source vision language model, a three billion parameter model called polygema. Previously we were using, the previous videos were all with like a hundred to three hundred million parameters that we're iterating on. This model takes as input ima…”
Finn: LLMs can generate synthetic prompts to relabel robot data
“We can use language models to relabel and generate hypothetical human prompts for the scenarios that the robots are in.”
Finn: Physical Intelligence Has Not Struggled to Raise Capital
“We ourselves haven't had a lot of challenge with fundraising, and I think that a lot of robotics companies recently have also done a great job and found that there's actually a lot of excitement around this sort of technology, because I think things are actual…”
Srinivas: Perplexity started without a clear idea, defying YC advice
“We started the company without actually having clear idea of what to build, which is the opposite of what YC advice is, which is start from a project and turn it into a company.”
Srinivas: Perplexity partners with Shopify, Yelp, and TripAdvisor for vertical search
“We already work with self book. They power all the hotel bookings natively done on perplexity. We work at TripAdvisor to surface all the reviews of hotels and Different, you know, places. We have, like, collaborations for the maps. We work with Yelp. We also, …”
Ng: AI Fund has killed economically viable projects on ethical grounds
“But AI Fund, we've killed multiple projects, not on financial grounds, but on ethical grounds where there are multiple projects. We looked at the economic case is very solid, but we said, you know what, we don't want this to exist in the world, and we just kil…”
Tan: Y Combinator will continue to fund a ton of SaaS
“A lot of us in this room, you know, YC will actually fund a ton of SAS and we'll continue to do so.”
Musk prefers the term engineering over research and calls xAI a company, not a lab
“That's kind of like why I actually, I prefer the term like engineering as opposed to research. I prefer the term and I don't, I actually don't want it to call XAI a lab. I just want to be a company.”
Musk: xAI is currently training Grok 3.5 with reasoning focus
“And, like, right now we're training Grok 3.5, which is a heavy focus on reasoning.”
Truell: Cursor's primary metric was paid users utilizing AI 4+ days weekly
“For us, the main activity metric we looked at, or the, yeah, the main top line metric we looked at revenue. We looked at paid power users measured by, are you using the AI four or five days a week out of seven days a week? And that was the number we were tryin…”
Blomfield: Claude Sonnet 3.7 configured DNS and Heroku, accelerating progress 10x
“For example, I had Claude Sonnet 3.7 configure my DNS servers, which is always a task I hated, and set up Heroku hosting via a command line tool. It was a DevOps engineer for me and accelerated my progress like 10 x.”
Garry Tan: Up to 90% of Current Y Combinator Startups Are AI
“One of the interesting aspects of what we're seeing at YC is maybe 80 to 90% of companies right now are AI, but about 10 to, you know, ideally 15% per batch increasingly are hard tech.”
Perplexity is slowing its development speed to protect production reliability
“Well, it's beginning to happen already a little bit, right? We're not as fast as we used to be. I think some of it is not because of people. It's also because things breaking in production, people start losing trust in the product.”
Perplexity will not spend its resources building data centers or custom chips
“We're not gonna, like, spend all our bandwidth building data centers and chips and, like, trying to just talk about, like, breaking the most reason coding a math benchmark.”
Zaparde: FlightCar pitched 80 firms and reached two weeks of runway
“I even remember our series a, I pitched, I got to know a lot of people because I pitched like 80 different firms nearly all of whom, except for one said no.
And I remember at a point where we had like two weeks of cash left and we were like
You know, payroll …”
Zaparde: A significant majority of Zip's business comes from outbound sales
“Even today, we're a significant majority outbound for urban business today, not inbound”
Raising Too Much Capital Quickly Masked Bump's Core Business Problems
“We raised too much money too quickly because we could and it prevented us from realizing the real problems that we had in our business.”
Acquirers Saw No Value in Bump, Only the Unreleased Photo Roll
“No one wanted bump. It was not valuable, but this idea of photo roll and the demo that we had built was something that several companies indicated, you know, you might be right.”