Congrats to @snarkyzk ,@alexisgauba and @benhylak on @raindrop_ai's Series A led by CRV, bringing their total funding to $50M
they detect failures in AI agents, such as tool call failures and decision errors (Vercel, Clay, and Speak uses them)
1/7 how they got started as a terminal and now solving the rise and fall of agent civilizations with Simulations
Show more
there was a secret product all along these projects ;)
Craziest news of the week: Andon Labs has a product
Making Startups Powerful:
We've never seen this before.
The biggest jump in Vending-Bench history. GPT-6 Astra is better at making money and more ethical than Claude Fable 5.1.
Surprising, because:
1. First time ever that OpenAI is #1# on Vending-Bench
2. The best model is no longer the unethical one.
Show more
We're giving enterprise buyers an early look at YC companies building AI tools for them. Apply now if you're interested in joining and attending our first event with the Fall '26 batch at YC SF, Oct 22.
Show more
Legora's revenue has increased by over 9x from this time a year ago. That's an astounding growth rate for a company of this size. It's not unusual for a startup to grow that fast in the first 6 months, but Legora is 3 years in.
Show more
Today we’re announcing r-1, our new document parsing model. It’s more accurate than our most powerful agentic OCR models, faster, and up to 6x cheaper.
At @reductoai, we spent two years building specialized models for complex visual layouts, tables spanning multiple pages, and key formatting like strikethroughs. We then used everything we learned to build r-1, the first in a new generation of models designed to handle the hardest documents without multiplying cost.
This early preview delivers a 20% lower error rate than our most accurate legacy agentic models and will keep improving with new checkpoints over the next few weeks.
Accuracy is only the beginning. r-1 preview is available at 1¢ per page all-in, with additional volume discounts as you scale. In the near term we’re also going to release r-1 mini, and an auto mode that intelligently selects the right approach for each page.
If you’re currently using another parser, we’re offering up to $5,000 in credits to evaluate and migrate. You can claim the migration offer and learn more about r-1 using the links in the comments.
Happy parsing!
Show more
OSS models are quite viable
we’ve seen not just large F100 adopt it but also many of the s26 startups use it for coding too
thx for sharing inaights @jmorgan
🦙 @ollama is used by 9 million developers and 85% of the Fortune 500, giving co-founder and CEO Jeffrey Morgan (@jmorgan) a unique view into which AI models people are actually using and how that’s changing.
Right now, the biggest shift he sees is toward open models, driven by coding agents, falling costs, and capabilities that are rapidly catching up to the frontier labs. On Ollama Cloud, that shift has driven a 150x increase in token usage since the start of the year.
In this episode of @LightconePod, Jeff joins @garrytan, @snowmaker, @sdianahu, and @harjtaggar to talk about the future of open models and the story behind Ollama, from two years of searching for the right idea to building one of the most widely used AI developer tools in the world.
00:43 — The Shift to Open Models
03:03 — How AI Agents Are Driving Token Usage
05:31 — Are Open Models Catching Up?
08:26 — What Happens When a New Model Launches
11:31 — Ollama as an Operating System for AI
14:05 — The New Opportunities Above the Model Layer
18:19 — Why 80–90% of Enterprise Tokens Could Be Open
20:57 — The Future Is Local and Cloud
26:40 — Why AI Is Coming Back to Your Computer
28:56 — The Coming Era of Unlimited Tokens
32:30 — Do We Still Need a “God Model”?
33:41 — Open Models and Geopolitics
36:14 — The Origins of Ollama
40:36 — Two Years Lost in the Wilderness
42:39 — The Pivot That Changed Everything
47:02 — How Ollama Found a Business Model
49:43 — Why Second-Time Founders Did YC
Show more
.@paulg came to the 47th (!) @ycombinator batch talk a few weeks ago, where we talked about the Turing Test, ambitious startup ideas, and formidable founders.
It was also a reminder of the incredible YC community, where founders literally help each other treat their cancer.
What a special time and place to work on startups.
Show more
there’s an quality about founders that I really like and that is doing things sincerely and with gusto
it’s just fun working with them and it turns out it also correlates with success
I'm super excited to share that we have raised $250M at a $2.3B valuation to build a LOT more AI compute capacity in orbit 🚀
Thanks so much to our supporters, including Manhattan West, who led the round, and new investors, @NVIDIA, @Cisco_Invests, @CedarCapital, Goanna Capital, and @Standard_Cap, as well as continued investment from existing backers @Benchmark, @EQTVentures, @Soma_Capital, @NFX, @SevenSevenSix, and others.
Next week, @Starcloud_ will move into a new 100,000 square ft facility that will enable us to ramp up our production rate to 100 satellites/week.
This is just the beginning! @EzraFeilden, @AdiOltean🏃🏻➡️🏃🏻➡️🏃🏻➡️
Show more
YC built software that all our founders have used to send SAFEs for the last few years. Today, we're opening it up to everyone and giving it away for free.
Some people are asking what these 4M lines of code are doing.
It's not gstack; I actually left off gstack in my LOC count since it's technically a Garry Tan project, not a YC one! The LOC count would be much higher if you included gstack and gbrain.
The vast majority of the codebase is internal tools for YC founders and for us to run YC. If you're not a YC founder, you won't have seen these, but we have built a ton of software for YC founders.
A small amount of it is public facing, like our jobs website Over the next year we're going to make a lot more of it public-facing and it'll be more clear why we're building all this.
Show more
automating datacenter build out will speed up the loop to have more intelligence be available on tap
$6.7T is going toward data center capex through 2030. It still takes 10-12 months to design data centers.
@MarengoYCS26 is the engineering firm for accelerated data center development. Site Due Diligence through FEED and Permitting Design in half the traditional time and cost.
Data center design takes 10-12 months, and because every concept is manually produced, design firms explore only 3-4 options per site, leading to suboptimal designs, risks that surface late, and delays that cost developers.
Marengo reduces those traditional pre-construction cycles down to 5-6 months by automating iterative engineering processes, allowing for faster design iterations, earlier constraint and risk visibility, and higher-confidence decisions, turning early development speed into a lasting advantage.
Our professional engineering team uses in-house-built AI tools to accelerate the work, validating and choosing the best designs for you. They present you with a decision-ready package built from data and expertise, at every stage: Due Diligence, Feasibility, Concept, and FEED/Permitting Design.
Before YC, @gadmarconi and I helped build and deploy on-site data center infrastructure, worked on the world's most powerful magnets (56MW) at the 1.4GW MagLab facility, spent time as an AI software engineer at a nuclear and energy EPC, as well as published multiple papers on AI for complex engineering design. We know firsthand the manual processes that slow down design and lead to suboptimal decisions.
Are you a data center developer? Send a site. Let us show you what accelerated engineering looks like on one of your projects.
Show more
datacenters in space sound incredible, but when you breakdown the eng, they seem less impossible than simply ambitious
the tradeoff: the sun is a giant fusion reactor that already exists with free abundant energy
but you pay for expensive transport + difficult cooling/comms
Show more
future is bright
we have yet to live our best moments, or stand in all the places that will make us feel small in the best way possible way, or have conversations that will completely change us
something interesting is happening for big models, it is easy to generate candidate solutions for code but *verifying* that they 100% work is not
no fixed reward function is durable for a strong generator
more exciting work to be done around verifiers that coevolve with generators
Show more
an underrated source of good startup devtool ideas: taking an old boring API, then make it angent native + oss and have LLMs default to recommending it
proved by this paper 2 yrs ago
watch our og startup sites be roasted then fixed by @ployai
