YC Startup School 2026 rehearsal at Chase Center
Welcome to San Francisco, future founders!
Tomorrow and this whole weekend are going to be 🔥
raghav
I just realized we could make SF better than Miami or Dubai, and now i’m pissed.
Sebastian Caliri: They pay nearly no property tax because of Prop 13, while you pay for their healthcare + other benefits you'll never enjoy.
Build the damn houses over Marina Safeway. Younger people in SF need a shot at family formation that doesn't hinge on getting liquid from OAI / Anthropic.
Vivian Midha Shen
T-1 until YC startup school!! Get ready for some BTS content
Elena
NOTES ON THE ODYSSEY (or…it was good, actually)
-something to keep in mind about all of christopher nolan’s films and heroes is how nonlinear they are, and that he seems to believe that great men don’t occupies a single fixed point in space and time (THE DARK KNIGHT / batman is maybe the one exception to this, for reasons I’ll explore later). his early thrillers like FOLLOWING and MEMENTO are about men trying to situate themselves in the timeline/the wreckage that ensues when they’re unable to. or recall the opening sequence of OPPENHEIMER which cycles through four principle layers of time (the 1954 oppenheimer security hearing, the 1959 strauss senate confirmation hearing, oppenheimer’s early college life, and strauss’ memory of meeting oppenheimer at princeton in 1947, plus quasi-prophetic imagery like the nuclear fireball, particles, etc). basically, most of nolan’s films are recursive loops that just iterate on themselves, but all the key themes/variations get revealed in the opening sequence.
-the great/notable men are constantly projecting themselves forward / backward in time, they exist everywhere, they (or their gods) have everything planned already.
-with this in mind, the odyssey lends itself perfectly to a nolan adaption. the book/text form of the odyssey is a nested narrative of 1) poet invoking the muse 2) action opens with the gods who discuss odysseus, then athena enters the human world and kicks off the telemachus plot which is ~20 years after odysseus left ithaca 3) penelope and telemachus refer back to odysseus’ departure/absence 4) a bard sings about the greek’s ill-fated return from troy. nolan pretty faithfully follows this with the bard sequence leading into the suitors leading into meeting telemachus/penelope leading into flashback to telemachus’ earlier childhood leading back to odysseus choosing to go to war/leave behind his family (and of course, these are all iterated through/supercut nonlinearly).
-in other words, the first 15-20 mins of oppenheimer map pretty much 1<>1 with the first 15-20 mins of the odyssey. both films follow this nested narrative structure (i actually leaned over to the person i was watching with and said “holy shit this is basically the opening of oppenheimer but ancient greece)
-something interesting about this trope of timeline-rupturing in most of nolan’s films is that it isn’t really about self-evaluation of one’s legacy. recall that (for example) oppenheimer and the odyssey only have one scene of explicit “remorse” each - oppenheimer’s hallucination in the auditorium where he imagines everyone as victims of the atomic bomb and the scene toward the end where odysseus is haunted by the sacking of troy / attacking innocent women / desecrating the athena statue etc). okay, caveat you could argue that the scene w/ heisenberg at the dinner was also a “regret” scene but for the sake of argument let’s just believe what im saying.
-to restate the above: nolan isn’t super interested in the subjective state of great men who seem to exist in the past/present/future all at once (like i said, he really just has one “subjective” scene each. fwiw after watching the film i also really disagree with that emily wilson clip going around that “this odysseus is super traumatized and introspective” or whatever she said). nolan is really just exploring an ontological fact of great men: that they are timely and timeless.
-all of the above said, im not 100% sure i liked matt damon’s performance. damon performs like a postmodern odysseus who’s already read the odyssey (or at least heard travis scott reciting it to him). he’s too knowing and too ironic. it feels too modern, he has iphone face even with that beard, etc. eg the scene in circe’s house where he clearly knows EXACTLY what’s going on but feigns obliviousness for comic effect just felt tonally off. if i had one complaint about the film it was in moments where damon plays stuff with too much deadpan 4th wall breaking effect.
Work life harmony
Nadia I: Had to make a quick fix at the ballgame - no laptop required. ⚾️ 👌🏼Thanks for the updates @okayzade @victoriakimse @Replit
If you haven’t used Replit for a while, you’re in for a big surprise.
Gene Lee: I have to hand it to the @replit team for the advancements made in the past year. I'm actually enjoying the building experience now.
Sorry I'm so excited... but, if you thought my Fable demos were cool, this one is just like 10 steps above
I don't think I'm sleeping tonight (and I've been sick all week so this is wonderful timing)
Matt Shumer: Okay I may have been extremely wrong about Opus 5...
Holy fuck, this demo is going to blow the entirety of AI Twitter's minds tomorrow
"How fast a country adopted new technology over the last 200 years accounts for at least 25% of why some nations are rich and others aren’t today. The Ottomans got the printing press eventually but eventually cost them a lot."
Julia Willemyns: I've been spending my weekends thinking about AI populism.
Sam Altman's home was attacked. A councilman found bullet holes in his wall and a note reading 'No Data Centers.' Political assassinations are in, which is hardly a good sign.
AI is coming for the credentialed class.
Will Anthropic sign?
If you work at Anthropic worth asking your leadership to sign or make their position clear: Are they for banning open weight models?
rat king 🐀: wow. openai decides to sign on to the Nvidia/Microsoft-backed open letter to regulators supporting open source
incredible
stories like this should be told more
Polymarket: JUST IN: ChatGPT surfaces a crucial mutation in a woman’s old biopsy records, helping doctors determine her terminal glioblastoma diagnosis was incorrect.
Greg Brockman
stories like this should be told more
Polymarket: JUST IN: ChatGPT surfaces a crucial mutation in a woman’s old biopsy records, helping doctors determine her terminal glioblastoma diagnosis was incorrect.
martin_casado
Wonderful to have a broad base acknowledgement of the need for open weights. We are delighted to support as signatories.
Open weights is the key to ensuring security, research, innovation and competition.
And let's not confuse misappropriation with openness.
Jensen Huang: For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
Raphael Schaad
POV: evening before the storm on Center Court — we rented out the entire Chase Center and are hosting thousands of the most ambitious builders in the world.
Welcome to @ycombinator Startup School 2026 🧡
Speakers include Jensen Huang, Sam Altman, and Jeff Dean. There’s also small group sessions with YC partners (yours truly included) where attendees can ask anything about startups. Going to be an awesome weekend in SF!
#sus2026
It was always perplexing to me that VCs passed on Etched in the early rounds. Glad they woke up… and early believers get less dilution 🙂
Etched: We’ve raised $300M in Series C funding at a $10.3B valuation from Sequoia, Andreessen Horowitz, Jane Street, Argo, and SK Hynix.
Our mission is to run the world's inference. This round accelerates production of our inference clusters.
We've opened an 80,000-sqft, 10-MW facility
Luther Lowe
I’m so proud of @weijia ❤️
Curtis Houck: CBS’s @Weijia Jiang at the #WHCD on her upbringing in rural WV: “I’ve been thinking a lot about my own journey to this podium tonight. I grew up in rural West Virginia, in a town at the foothills of the Allegheny Mountains, a town called Buchanan, Population 5,500. Those images
Pets make building fun
OpenAI Developers: Your Pet is ready to meet other builders.
On ChatGPT web, create a shareable link for your custom pet and send it to a friend to adopt.
Luther Lowe
I had a chance to chat with @howardlutnick tonight at the White House Correspondents’ Dinner, and he said directly to me: “This White House will protect open source AI.”
Donald Trump Portfolio Tracker
January: Trump’s account buys $1–5 million of Nvidia in a single trade.
April: He bans Nvidia from selling its chips to China. The stock craters, Nvidia takes a $4.5 billion write-down.
Through the spring: his accounts keep buying the dip, over a dozen Nvidia purchases.
August: He reverses his own ban. Nvidia can sell to China again, if it pays the U.S. government 15% of the revenue. A deal he cut personally.
August 18: his account buys $5–25 million of Nvidia, the single biggest stock purchase of his entire year.
December: He greenlights the even more powerful H200 chips for China, now at a 25% cut.
The President banned the stock, bought the crash, un-banned the stock, then made his biggest purchase of the year.
He controls the switch. He owns the stock. He flips the switch.
Theo - t3.gg
Re @mattshumer_ Just saw the demo.
What the fuck.
Tomorrow.
Turn post notifications on :)
Theo - t3.gg: @mattshumer_ Just saw the demo.
What the fuck.
Gandalv
America’s Quiet Exodus!
Neil deGrasse Tyson has been unusually blunt lately: he compared federal science cuts to an act of war, arguing that if a foreign rival gutted American research budgets this way, Washington would call it sabotage. Instead, it’s self-inflicted.
The numbers back him up. NSF grant awards recently hit their slowest pace in decades, with physics funding down over 80 percent and biology nearly halved. A Nature poll found roughly three-quarters of US researchers now considering leaving the country entirely.
The beneficiaries are obvious. France, Canada, and Australia are actively recruiting displaced American talent, dangling stable funding and political neutrality. China, meanwhile, has overtaken the US in total research spending, a shift analysts warn could cost America a trillion dollars in lost competitiveness over time.
Tyson’s warning isn’t nostalgia. It’s a balance-sheet argument: brains, once they leave, rarely come back.
https://youtu.be/dYiUI6y1nKg?is=gZ0jdubmLC9mtCt
ChatGPT Work for using websites which require login:
OpenAI Developers: Your ChatGPT Work agent can now use websites that require you to sign in.
Take over the cloud browser to log in, then let your agent continue the task. Your login persists across sessions, so you only have to sign in once.
Christopher Nguyen ⽗
The Gemma series is an amazing set of highly performant open-weight models. They have proven extremely effective in industrial settings where site-deployed agents need exactly this as a base for domain-specific fine-tuning that even Fable-class models cannot match in accuracy.
Presumably the Gemma team has been excellent at “distilling” from the Gemini-class models to get such capabilities.
This is a $25T category. Remember, the “frontier” is not spherical. It is extremely jagged, and it isn’t just among the trillion-parameter group.
Clément Farabet: Very much agree with this letter.
3 years ago we started Gemma aligned with this vision, that building a thriving open weight model ecosystem was critical and likely a huge component of our future industry.
Today we continue to invest and push Gemma to be maximally useful for
Miguel de Icaza ᯅ🍉
It’s giving pull the ladder behind you energy.
Anthropic using open source to power their business ok. The world using open weights, not under Julian’s watch!
Julian Schrittwieser: And @satyanadella coming out for open source too, can’t wait for the open sourcing of Windows and MS Office!
YC Startup School is for many thousands of technical young people people the very first time they step foot in San Francisco
Welcome to the place dreams are born!
Pasha Khoshkebari: Flight to YC
@lucashjin and 60 more of us
I love how their core canon is the Karen Hao book that spread all the lies about water in AI data centers (it was so wrong they had to retract that claim, and yet it is still repeated everywhere)
sarah 😾: Proof SF is a bubble. I wonder if there were protests and flyers when cars surpassed the horse and buggy and the printing press was replaced by digital media. Innovation inevitably leaves people behind.
Jason is a DevEx engineer at OpenAI who wrote the official guidebook on how to use Codex for work.
In tomorrow’s episode, he showed me his complete Codex work system, including how to:
→ Run a chief of staff across Slack and email
→ Turn past sessions into new skills and workflows
→ Build sites to learn anything, like the drums
📌 Subscribe to get the full episode tomorrow: https://www.youtube.com/@PeterYangYT?sub_confirmation=1
Peter Yang
Jason is a DevEx engineer at OpenAI who wrote the official guidebook on how to use Codex for work.
In tomorrow’s episode, he showed me his complete Codex work system, including how to:
→ Run a chief of staff across Slack and email
→ Turn past sessions into new skills and workflows
→ Build sites to learn anything, like the drums
📌 Subscribe to get the full episode tomorrow: https://www.youtube.com/@PeterYangYT?sub_confirmation=1
Y Combinator
gm startup school
Suhail
Re Supporting open source does not mean you must open source. I know you know that. So, if this is how you're making yourselves feel better in Anthropic Slack, this is quite concerning.
You don't need to open source anything but spending 10s of millions lobbying and doomsdaying people's open source work is a major problem I will never stop fighting your company on.
Michael Dell 🇺🇸
What @tobi is doing here is remarkable: running GLM-5.2, a 753B-parameter AI model, locally on a Dell Pro Max with GB300 at 40 tokens per second.
No data center. No API. No cloud.
Unmetered intelligence. Frontier-scale AI is becoming personal.🚀
tobi lutke: Thank you @nvidia for the hookup on this DGX Workstation. This will crunch a *lot* of high quality tokens here!
This thing is a total beast.
http://x.com/i/article/2080484722204131328
Ankit Gupta
My introduction to startups (that I didn’t act on for 4 years) was in my freshmen year of college in 2013 when I attended startup school.
Can’t wait for an incredible group of students to share that experience (except 100x better than it used to be)
Y Combinator: gm startup school
Peter Lewis
No one realizes how quickly our world is going to change…
Frontier AI models are crafting games like you see below in a single pass from a single prompt.
Buckle up, the future is EXCITING!!!
Matt Shumer: Claude Opus 5 one-shotted this game.
EVERYTHING you see in this demo is custom code... not a single external asset was used.
AI games are going to be amazing.
(sound on)
Zach Graves 🍞🎪
Powerful energy here in SF @garrytan @ycombinator
max
YC Startup School 2026 is live!
IMO it really sucks for those who have put their lives into game dev... but ppl burying their heads in the sand are doing themselves (and others) a disservice.
The sooner people are aware of what's coming, the sooner they can position themselves well.
Chris: As I’ve been saying for the past 6 months the end of indie gaming studios is nigh. Soon it will be in the people’s hands.
This is why I love Matt as well, he doesn’t produce slop. He pushes AI models to the max and makes all other demos look like absolute slop.
Boris Jabes
Jensen: we always approached problems with the attitude of “how hard could it be?”
Evergreen startup advice from #ycsus26
Brad Flora
#YCStartupSchool dropping in to say hello to a builder doing an office hour with AI Brad. He said AI me was asking good questions about his GTM motion!
Susan Zhang
open weight models are indeed Very Useful! i think we all benefit from extra free data floating around in the ecosystem, even when we're totally not intentionally taking advantage of these open models at all, they will still show up and help improve our systems, crazy useful!
Julian Schrittwieser: @satyanadella funny to see people jump to the conclusion I must want to ban open weight models.. I actually think open models can be very useful!
But it’s interesting how some historically extremely anti-open source companies are suddenly all in favor of openness 🤔
Ti Morse
My first interview with @sama, Co-Founder of @OpenAI.
0:04 How to start a startup
3:30 Trusting exponentials
4:57 Operating in chaotic environments
6:12 Learning to enjoy painful experiences
8:03 Creating abundant intelligence
11:15 Keeping core suppliers on OpenAI’s timelines
12:10 Invention of the joint-stock company
15:30 The best CEOs aren’t sociopaths
16:46 We are in the singularity
18:09 AI authoritarianism vs liberty
19:24 Texting 300-400 people a day
20:32 Having a small number of deep beliefs about the future
21:41 Critical path
22:25 Thinking about what’s next
23:38 Getting on planes in marginal situations
28:16 Buying lots of compute
30:51 Ambition
34:00 Google shouldn’t have let OpenAI survive
37:54 Having his life shot through a cannon after the launch of ChatGPT
41:20 The growth of Codex
42:02 The Death Star tweet
44:46 Status games and desire to be useful
47:57 Not being ambitious enough on compute investments
50:26 Execution
51:45 Ask for what you want
54:11 First few weeks of OpenAI
55:15 Shutting down Sora to focus on Codex
57:44 Designing beautiful products
59:01 Getting addicted to TikTok
1:01:03 Inventing a new device
1:02:53 Try to get better at your strengths
1:05:57 Masa is an n of 1
1:07:11 Real trends vs fake trends
Gokul Rajaram
Your Ceiling Is Delegation
@jess__yan , product lead for Claude Managed Agents, Anthropic, interviewed by @petergyang (Behind the Craft)
Summary: Jess Yan leads Claude Managed Agents at Anthropic, and her core claim is that agents have crossed from prompting loops into autonomous, long-running coworkers you can load with work overnight. How much you can delegate at once now caps your output more than your personal capacity does. The fastest path to value is making one person exponentially more powerful, then scaling from there.
1. From Loops To Coworkers. Agents stopped being prompting loops and became autonomous, self-recovering, long-running actors. Once they were just question-and-answer loops; now they hold permissions, access to internal tooling and sensitive data, memory, and the judgment to know when to keep going versus ask a human. The harness, meaning the scaffolding around the model that runs tools, calls memory, and decides when to pull a human in, is what turns raw token-sampling into a product. As the tasks get more ambitious, the harness gets thicker while the prompt stays simple.
2. The Ceiling Is Delegation. The new limit on what you get done is how much you can delegate at once. Yan puts it directly: the limits of what we can achieve will be based on how much we can delegate, more than on our personal capacities. The working model is to set tasks overnight and wake up to a resolved backlog and squashed bugs. Load agents up with work the way you would brief a partner.
3. Outcome Is The New Output. Structured outputs used to mean forcing rigid JSON and gluing it together by hand. Now the outcome itself is the spec: describe the rich, interactive thing you want and let the agent self-correct its way there. Yan calls this a tastemaker's assessment of what good looks like. You set the standard, the agent iterates to hit it, grinding a predictive model until it clears 90% accuracy. The skill shifts from writing formatting instructions to defining the standard.
4. Empower The Individual First. The mistake enterprises make is jumping straight to automating a 20-team, multi-month workflow. Yan's counter is to make one person exponentially more powerful first, giving them agents that do the design work or handle the dependency request they would have waited on a human for. You end up with a bunch of one-person startups inside a large company. Raise the ceiling on what individuals can ship alone, then go after the mega-processes.
5. Spin Up, Throw Away. Treat agents as disposable. Yan keeps a different agent running every couple of weeks, each suited to the task of the moment, and discards it when the need passes. One example: a 4,000-org waitlist full of duplicates and junk, where she built an agent with access to internal systems to parse it, score who was likely to convert, and pull people in daily, knowing the whole thing would be obsolete in a few weeks. It does not have to be productized. It has to work now.
6. Vibe Test Before You Eval. Get the agent into real users' hands before building formal evals. Yan says vibe testing is the most important first step. You outgrow it once you cannot aggregate vibe signals at scale, and that is when evals come in: binary pass/fail, LLM-as-judge scoring, and triggering evals that check a specific action actually fired. Do not burn weeks on eval scaffolding before anyone has touched the thing.
7. Agents Grading Agents. Let agents evaluate their own work, ideally in a separate context window to avoid bias. Claude Managed Agents can pull the eval loop directly into the session, so the agent grades its outputs and revises before a human ever looks. Yan's own loop mirrors what builders do in Claude Code: generate, have another agent judge, regenerate if the grade is bad. Self-correction inside the loop beats stringing together human checkpoints.
8. Always-On Agents. A real coworker does not wait to be asked. Yan runs agents that monitor Slack channels, summarize customer activity, and surface things on their own through triggered events and cron jobs, constantly refreshing context so they stay as current as she is. The two sources of power are the level of data access you grant and an interaction style she describes as humanlike and proactive. An always-on agent has to pop up at the right time, in the place the work already lives.
9. I Talk To Claude More Than Coworkers. Yan admits she talks to Claude more than her teammates, spending therapy time wrapping her head around thorny concepts before a meeting. The effect is that she arrives with a real opinion and baseline research already done, so conversations with colleagues start at a deeper level. Anthropic even tags Claude into API design reviews as a neutral judge when the team hits an impasse. The agent uplevels human conversations rather than replacing them.
10. SaaS Goes Hyper-Vertical. As models absorb broad domain expertise, the value moves to incredibly specific, niche use cases. Picture an accounting agent built just for solopreneurs. The durable shared asset becomes context patterns and task-orchestration patterns that travel across these narrow agents. The products that survive are the ones that meet users exactly where their workflow already lives.
11. Software Lives In Chat And Code. The interface is collapsing into chat and the coding environment. Yan's point: you bring your personal agent, loaded with your contacts and context, and let it go deal with someone else's agent, instead of filling out forms on a website. Everyone is becoming an engineering team, so work increasingly lives in Claude Code and in chat. Build where the agents already are.
britton
jensen gave the best advice for CEOs about company building today at YC startup school
F1 teams build the car for the driver.
Founder is the driver.
win your race - build to win
(someone will clip the quote when the vid is up)
Aaron Epstein
“This is absolutely the single greatest time to start a company, I’m jealous of all of you” @JensenHuang at Startup School
Ahmed Karrar
Biggest takeaways from @JensenHuang CEO @nvidia at the @ycombinator startup school!!
> You don’t have to change your personality or behavior when you become CEO.
> You are building an F1 racer that you are going to drive so you need to build it the way you want to drive it.
> Systems understanding, awareness, and design is crucial (most of the low level things are going to be automated)!!
> Understand agent as AGENT is the new software!!
> AI automates task and dose not necessary eliminates jobs as the job of a person has a purpose hence why AI will create more jobs than taking jobs!
> The ChatGPT movement of robots happened already! Robots walking around, reinforcement learning happened years ago and it helped us open our thinking!!
> DONT BE INTROVERTED “took me until 2026 to have my first post on X”
Message he would send to his 18-20 year old self:
> Everyday day the world is changing and this is the best time to start a company!
> Don’t imagine how hard it can be just get going on working at it!!
> Believe in your ability to LEARN.
> It is going to be hard just have the believe that you can overcome TODAY.
RESILIENCE IS THE MOST IMPORTANT THING.
Truly grateful for this opportunity thanks @ycombinator!!
If you enjoyed this be sure to like and follow as I will make sure to do write all takeaways from the speakers today.
Packy McCormick
Anthropic is an ideologically captured organization seeking regulatory capture to protect both their business and their ideology.
Packy McCormick: AI is Oil, Not God
"We should not tax every American business by forcing them to pay the storyteller to protect them from the story."
https://www.notboring.co/p/ai-is-oil-not-god
Boris Jabes
Boris (from Claude Code): you should delete your system prompts and skills every 6 months and see what the new models do
I find them baffling even to me. Especially bad for the ADHD brain.
Ethan Mollick: Try to teach a novice Code or Codex and you will realize how baffling they are to many.
They assume you know model names & thinking levels & projects vs. folders & skills & plugins & connectors, etc.
Even just clicking "+" can lead to a flood of complexity. Much is undocumented
Jared Friedman
Doing a little Claude Coding while watching @bcherny at Startup School. It's such a treat to get meet the creator of a product you use every day.
Jared Friedman
Re Could not have done it without essential help from the awesome team at YC: you @eve_bouff, @raphaelschaad, @evansolomon, and @aaron_epstein all contributed key pieces.
Also couldn't have done it without the incredible products built by YC alumni that power the experience: @lina_colucci, @bryantchou, @sumanyu.
Just deployed a new chess engine; we're closing in on an estimated 1200 Elo.
The goal is to get it to 2000+, while keeping the constraints:
1. One small finetuned LLM (no custom pretraining or architecture)
2. The model has to produce the moves with no chess engine assistance
If you relax these constraints, it gets much easier.
Amjad Masad: If you want to play my chess engine (WIP): https://qwen-chess.replit.app/
It already seems to perform better than frontier models on chess.
It’s fine-tuned on 2M stockfish-labeled positions, then short GRPO RL pass.
Documentation/tutorial with all the experiments and annotated code.
Diana
just went on stage with @bcherny for startup school
and he had a special announcement for all 6k+ attendees: free claude max 20x accounts for 6 months!
thank you!!
YC Startup School 2026 just got going
3x bigger than last year
We made this event for first time very technical young people to come to SF and figure out if starting a company is for them
Let you in on a secret: it is
Paul Graham
For all the young founders who told me two or three years ago that they had to drop out of school to start companies at that moment or it would be too late:
Aaron Epstein: “This is absolutely the single greatest time to start a company, I’m jealous of all of you” @JensenHuang at Startup School
spor
game day in the center of technocapital
Sanjana Friedman
FOLLOW YOUR DREAMS
I started doing all my research with agent CLIs and the filesystem. A 📁 𝚛𝚎𝚜𝚎𝚊𝚛𝚌𝚑/ folder, an 𝙰𝙶𝙴𝙽𝚃𝚂.𝚖𝚍 describing the format I like and best practices. Then launch my agent and ask a question.
No fancy apps, knowledge graphs, UIs. The agent can be asked to find and correlate knowledge from previous sessions. Scales to infinity. The folder can be synced via iCloud, git, etc.
If I want to share some outcome, I can ask the agent to render a cool HTML report and deploy it to Vercel. The "software" that powers this is English in 𝙰𝙶𝙴𝙽𝚃𝚂.𝚖𝚍.
Jared Friedman
Deploying code live in the middle of startup school with @WhatKindOfKat to make the event better. YC really is just a big startup.
Grace Tang
“founder mode can last 34 years, from zero to 5 trillion” - @JensenHuang
might become a jensen huang stan ngl. thanks yc startup school, cool to see him talk!
Joe Hu
Personal AGI.
cognition buying poke is the kind of quick @ScottWu46 acquisition that makes more sense with each passing day
Phil 🍵: we installed claude tag and now someone on the team needs to tell it to shut up every 30 minutes
The (software) factory is the product. Your product is only as good as the agents you set up to autonomously maintain it.
This is what @elonmusk figured out about Tesla, and it's now true for the software world.
Jonathan Ouyang
I never thought I would see a stadium so full of entrepreneurs and builders. Shoutout @ycombinator
eve
“so why do you work at yc?”
Sammy Tourani
"A father built a digital brain out of 80,000 markdown files containing every seizure log, doctor's note, lab test, and medication his son ever had, and fed it into an Al to get answers.
Nobody would have ever built that for him. But he built it for himself.
One laptop. One Al agent.
That's the real power of Al."
- Garry Tan @ycombinator
My friend Kun gives some of the best analysis on models there is, read it below:
Kun Chen: opus 5 is a VERY interesting release for a few reasons
1. it showed that the general benchmarks we use today are almost completely useless now
opus 5 is nowhere near fable in practical use, not even close. anyone who’s used it meaningfully can tell this very quickly after a few
tax the poor
California YIMBYs do genuinely need to find a way to get rid of prop 13, it’s basically statewide rent control and subsidization of boomers by working age people
tax the poor: Get rid of their property tax/rent caps and watch them become the biggest YIMBYs on the planet
aiden
incredible breakout session from garry tan at startup school, transformed my entire point of view on ai agents.
for my homework tonight: creating my shelf 🫡
Zac Burton
Re @JensenHuang & @garrytan at YC Startup School;
“Controllability of agents will be the next breakthrough”
He’s referring to how a simple word change in an md file (for example) could significantly perturb agent actions.
Interesting area!
lmao @ClementDelangue doing the 🇳🇴
Era Qian 👩🚀🔎: 🎉 Today’s the day! About 500 open-source builders are coming together
Join us to support the Open Weights Movement. Bring friends and good vibes! Meet all the gladiators pushing the open AI frontier 💪🏻
Henry Zhang
Genuinely amazing talk.
http://Eve.dev is more fundamental than any other framework we’ve built.
It’s the genesis of your company. When you have a new idea, don’t think: “I’ll go to an agent and ad-hoc prompt.”
Think: how can I build the factory that will start, maintain and grow my idea
Guillermo Rauch: The (software) factory is the product. Your product is only as good as the agents you set up to autonomously maintain it.
This is what @elonmusk figured out about Tesla, and it's now true for the software world.
Thank you @JensenHuang
Your lessons are an incredible gift to the next generation of entrepreneurs
Jake Roggenbuck
Interesting to hear about @bscholl’s “Confusion List”, where he kept track of topics that he didn’t fully understanding at first principles. He’d try to take an item off that list each week, except this list kept growing, but he still found it useful.
Startup School is great so far! Looking forward to tomorrow.
@ycombinator @boomsupersonic
Legends @JensenHuang @bcherny @sdianahu
Startup School is so special
From photographer future founder at Startup School 2008 to this today
Look ma, I made it
Garry Tan: In 2008 I showed up to Y Combinator Startup School to take photos of speakers and learn how to start a startup.
I was an engineer and designer and had never started a company before.
britton
which way western man?
Alex Hernandez
Fantastic breakout session with @dflieb at @ycombinator #YCStartupSchool ‘26! Great advice, amazing co’s and founders. Super grateful to be a part of it.
Cristian Dinuta
spinoza would have loved gstack
Diana
thx @JeffDean! we ended on my favorite question
someone in this room will build something as foundational as MapReduce or the TPU
what problem do you hope they picked?
x.com
