←
Y Combinator27,903 posts
Y Combinator avatar

Y Combinator

@ycombinator

We help founders make something people want. Subscribe to our newsletter: yc.link/newsletter

◷ Joined February 2010▧ 4,365 Photos and videos
Y Combinator avatar
Y Combinator@ycombinator·Nov 14

At @spacexai, Lauren Tan (@poteto) and Roman Ugarte (@romanugarte_) are on the core team that took Grok Bot from first line of code to launch in seven weeks. It lets you spin up AI teammates with their own identities and logins, and they just keep working in the background. Their marketing team gave a bot its own ad budget. Lauren built an entire feature with her bot in one Slack thread without ever opening Cursor. They joined us on the @LightconePod to talk about what it’s actually like working next to bots every day, and how the hard part is shifting from doing the work to figuring out what to do and why. 03:36 — Giving Bots Real Jobs 06:49 — One Bot or a Team of Specialists 10:39 — Sharing Bots Across the Company 14:46 — Identity and Permissions for AI Coworkers 18:37 — Bot-to-Bot Collaboration and Trust 22:35 — The Workflow Behind 4,000 PRs a Month 26:00 — Teaching Agents to Write Code You Trust 29:01 — Building a Personal Software Factory 33:44 — The Future of AI Coworkers 36:47 — Rethinking the AI-Native Company 40:13 — Teaching Agents to Check Their Own Work 43:56 — How Much Code Review Is Enough? 47:33 — Codebase Guardrails That Save Tokens 52:20 — When Product Judgment Becomes the Bottleneck 56:29 — From Finance to AI Engineering

1:00:06
Ishaan avatar
⟳ Y Combinator retweeted
Ishaan@ishaan_jaff·13h

We’re launching agent evaluations in the @LiteLLM SDK for developers who want to compare models and harnesses on their own tasks. Before you had to write the eval loop, create temporary workspaces and collect the results in a spreadsheet. Now you can reuse your agent, define typed tasks with a success check, and call litellm.agent_eval(). You get a scorecard with pass rate, cost, tokens and duration.

Post attachment
Coris avatar
⟳ Y Combinator retweeted
Coris@CorisInc·Nov 16

The math adds up. The bank statement is still forged. Introducing 𝗙𝗼𝗿𝗴𝗲𝗖𝗵𝗲𝗰𝗸 by Coris: document checks that surface signs of tampering and inconsistencies, with evidence reviewers can investigate. Try the samples: forgecheck.app.coris.ai

0:38
Paul Graham avatar
⟳ Y Combinator retweeted
Paul Graham@paulg·18h

It's true. Startups are basically a series of disasters. Although YC's official motto is "Make something people want," we had two unofficial ones as well: "Never a dull moment" and "You can't make this stuff up."

When I first joined YC we had a running joke that “there's never a dull day” because however many disasters we'd help founders with, we were always discovering new types of disaster. 16 years later, there's still never a dull day.

View quoted post ↗
conor brennan-burke avatar
⟳ Y Combinator retweeted
conor brennan-burke@contextconor·Nov 15

no matter how big @Hyperspell becomes, I want us to always make time to try new products from early @ycombinator startups we’ve been some of the first customers for a few YC companies in recent batches, and they genuinely feel like an extension of our team we regularly have founders from other startups coming by our office, helping us solve problems, getting feedback on new features, or building something for our exact use case it feels like we have an army of experts across different domains helping our team punch way above our weight it completely changes the build vs buy equation. instead of building something in-house or buying an off-the-shelf product, you can have an entire startup building specifically for your needs of course, it’s a bit like investing. you need to bet on the right founders because the products are often super early, but when you get it right, it’s a game changer

Lucy Cai avatar
⟳ Y Combinator retweeted
Lucy Cai@lucyjcai·19h

We taught a house full of robots to throw a dinner party. Watch the robots run it live tomorrow (11/18) at 4pm PT. Livestream link in thread

0:57
Junaid avatar
⟳ Y Combinator retweeted
Junaid@thejunaidm·Nov 16

Introducing Poseidon: a state of the art cybersecurity specific model, designed for vulnerability research, finding, prove, and fix vulnerabilities in real software. Poseidon preforms: - 17.6% higher than GPT-6 Astra - 2.9% higher than Claude Fable 5.2 (CWE-Bench)

0:15
Aarush Yadav avatar
⟳ Y Combinator retweeted
Aarush Yadav@AarushYadav47·20h

A one-sided conversation. You can have a conversation silently with Subvocal (YC F26). I couldn’t hear Arnav, but the glasses could.

0:45
Sanchit monga avatar
⟳ Y Combinator retweeted
Sanchit monga@sanchitmonga22·13h

Introducing Wally Code: UNLIMITED coding at ₹1,000 a month for students. Same model, about 12x cheaper than @SarvamAI's API and 21x cheaper than @deepseek_ai's own API, for an agent running all month. Sarvam got picked to build India's own model and launched a per-token API. This is what India deserves, built with no grants and no subsidies.

Post attachment

Sarvam got picked by the government in April 2025 to build India's own foundation model. They got a ₹98.68 crore GPU subsidy for 4,096 H100s and raised $234M at a $1.5B valuation in June. Now they've opened Token Factory to everyone. 8 models on it. 1 is theirs. 6 are Chinese open weights (DeepSeek, Kimi, GLM) and 1 is Google's Gemma. So India's sovereign AI lab is mostly serving Chinese models from Indian GPUs. That's a hosting business, and plenty of companies already run one. I wanted Sarvam to be the lab that ships frontier research out of India. After that much capital and subsidized compute, this is not it.

View quoted post ↗
Shire avatar
⟳ Y Combinator retweeted
Shire@ShireIntel·13h

We are building a POS that runs your restaurant for you. Restaurants run on thousands of tiny decisions every shift. Who gets seated where? Which tables need attention? When do servers need help? Today, all of this depends on managers constantly watching the floor, coordinating staff, and catching problems before they snowball. Traditional POS systems record what happens in a restaurant. Shire (YC F26) is building one that actually runs it. Reach out to see how Shire can transform your restaurant: founders@shireintelligence.com

1:32
Kevin Miao avatar
⟳ Y Combinator retweeted
Kevin Miao@KJHMiao·21h

Today, we’re announcing @perceptual_labs. We believe weight-space programming will fundamentally change how we build, serve and shape intelligence: shifting from training models to engineering them, and eventually toward models that can improve themselves. When I started in interpretability in 2019, a famous researcher I admire asked me: “Why do we need to understand models if all we care about is that they work?” I’ve been trying to answer that ever since. We spend so little time understanding what models learn and so much brute-forcing data and compute.  To me, that isn’t intelligence.  My applied research at Apple and Berkeley has been about the opposite: nimble systems that learn from little data and behave in targeted ways. Understanding models is how we make them better and how we make them safe. With recent advances in weight-space programming and alignment, I think the time is finally right. Join the waitlist and learn more: perceptuallabs.com

0:16
Aarush Agrawal avatar
⟳ Y Combinator retweeted
Aarush Agrawal@codingwithaaru·14h

Today we’re launching Contour (YC F26): AI-native enviromental permitting and compliance for America’s buildout. America can’t build faster than it can permit. We’re buying the firms that write the permits, and running them on AI.

1:06
Adi Mashiach, MD avatar
⟳ Y Combinator retweeted
Adi Mashiach, MD@AdiMashiach·21h

Today we're excited to share Bonbon, a vertical foundation model for biology with emergent biological capabilities (#NeurIPS2026 Oral). Bonbon independently learned the hidden relationships governing molecular biology, opening a path to understanding biology as a complex system. 1/9

0:45
TRISHALA avatar
⟳ Y Combinator retweeted
TRISHALA@hitrishala·16h

Welcome to the real world. Today we’re introducing FrontierEE Bench, the first benchmark to test frontier models on electrical engineering tasks at a physical lab bench. Real-world engineering benchmarks should involve real hardware. At our lab, models control instrumentation, run experiments on physical hardware, and decide what to test next. The best models scored 36.7% across a range of EE domains including mixed-signal circuits, power systems, & electromechanics.

0:05
Amanda avatar
⟳ Y Combinator retweeted
Amanda@amandarecall·21h

We made an actual movie during a call. In this demo, my bot brought a video generation agent into the call, which then made the cute penguin video. No coding, clicking - nothing except talking to the bot. Anyone on the call can direct it and watch the changes. There are so many things you can have your agent do. Immediately. During the meeting. You can build this with Recall.ai’s Output Media, which lets agents stream audio and video into meetings. What do you want your agents to get done for you?

1:54
Streamline Systems Incorporated avatar
⟳ Y Combinator retweeted
Streamline Systems Incorporated@No_GPS_Needed·16h

Streamline Systems (YC F26) is building the replacement for GPS. GPS signals are easy to jam and spoof. When GPS is denied, drones fail or revert to manual piloting. Existing alternatives are too expensive for modern warfare. Reach out: charlie@streamlinesystems.ltd

0:51