Our @ycombinator x @GoogleDeepMind Startups Day w/ @demishassabis and @garrytan was a major life highlight.
For those who weren't able to join, I put together a few of the top themes, quotes and ideas.
https://aistudio.google.com/learn/yc-gdm-startups-day-building-agi-era
Of all the fantastic quotes this one from Demis stood out:
“Life's very short...you might as well put your life force into something that will really make a difference.”
I hope this is helpful and inspiring for you as it is for me.

dm ya boi for health coaching if you have wha…

A reminder: model outputs are not IP. It’s just non-copyright text. No different than an AI image!
Free to distill to your pleasing. After all, it was humanity’s data to begin with.

Jensen Huang on "distillation"
On his new interview with axios, he was asked this question
"Should open source model companies be allowed to distill closed models"
"Distillation—learning from AI, learning from other people, and learning from other sources of knowledge, is fundamental to intelligence.
We are constantly learning from other people. I am learning from you through the questions you are asking, and you are learning from me. All day long, we are learning from one another. AI also has to learn from something.
The original AI models, whether they were open or closed, were trained on previously created knowledge from the internet. Now, AI is generating more content than humans. In a few more years, the internet could be 99% AI-generated content, and that content will have been created by some form of AI.
As a result, AI systems will constantly be distilling knowledge and intelligence from other AI systems. The fact that AI can learn is a good thing. We want AI systems to be intelligent because a smarter AI can also be a safer AI."
----
From "Axios" YouTube channel, (full video link in comment)

When I was in Paris recently, I had breakfast with a deep tech founder who was debating wether to move to SF. He wanted to improve his odds at fundraising.
That's the wrong framing. Capital matters but the biggest benefit in coming to SF is to raise your ambitions. You are the average of the people you surround yourself with. In SF, the people surrounding you are attempting things that would sound insane in Paris.
I've watched it happen to founder after founder, and it happened to me too. Put yourself somewhere you're no longer the most ambitious person in the room!

AI Engineer Paris is back!
After an incredible first edition, we’re excited to announce AI Engineer Paris 2026, this time hosted by our friends at @MistralAI.
 🎤 CFP is open. Submit your talk by July 31.
🎟️ Super Early Bird tickets are sold out! Early Bird tickets are now available.
Get your ticket today and see you in Paris! 🇫🇷
http://ai.engineer/paris

NVIDIA, Palantir, Replit, Microsoft, Crowdstrike, Dell and others send a strong message to congress to keep open access to open weights models.
"Our AI leadership will be judged not by one frontier AI model, but whether the United States builds a strong, open ecosystem that diffuses into every sector."
"Open weights also strengthen competition and competition is what keeps the gains of AI broadly shared rather than concentrated in a few hands."
"Openness may be one of the most important paths to AI safety and security."

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.
The world needs both frontier closed models and frontier open models.
https://images.nvidia.com/pdf/Open-Weights-and-American-AI-Leadership.pdf

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.
The world needs both frontier closed models and frontier open models.
https://images.nvidia.com/pdf/Open-Weights-and-American-AI-Leadership.pdf

Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. https://www.microsoft.com/en-us/corporate-responsibility/topics/open-weight/

Wow.
NVIDIA just dropped the clearest, strongest letter yet on open-weight models — co-signed by Microsoft, Palantir, ServiceNow, Box, me and a growing list of real builders.
Read it. Internalize it. Share it.
Open weights are not a risk to American leadership.

They are American leadership.
They expand access so startups, universities, hospitals, factories and main-street businesses can actually use frontier capabilities instead of renting them from three closed labs.

They create real competition.
They give companies sovereignty over their own models and data.

They strengthen security through transparency instead of hoping a handful of black-box providers never get breached or decide to gatekeep.
The letter says it perfectly:

“Policymakers have an important opportunity to act… keeping the frontier plural by avoiding premature restrictions on open models that stifle competition or drive innovation overseas.”
Exactly.
Which is why we have to push back hard and now on Dario Amodei and Sam Altman.
Both have spent the last two years lobbying Washington for exactly the kind of restrictions this letter warns against — framing open weights as an existential danger while their own closed models sit behind high walls and high prices.

Their solution always somehow ends with fewer competitors, higher barriers, and more control concentrated in the same few labs that already dominate.
That is not safety.

That is moat protection dressed up as safety.
NVIDIA and its partners just put the opposite case on the table in public: openness is how America wins. Diffusion is how America wins. A plural frontier is how America wins.
If we let the closed-lab lobby write the rules, the only “open” models left will be the ones coming from overseas.
The letter is out.

The case is clear.

Push back now.
Open weights.
American leadership.
No premature restrictions.
NONE!

Since the start of the year, @OpenCode (YC W21) — an open source alternative to Claude Code and Codex that works with any model — exploded to 4.6 million weekly active users, 13 million monthly actives, and roughly $40M in annualized revenue.
In this episode of The Lightcone, @harjtaggar, @snowmaker, and @sdianahu talk with Jay V (@jayair), OpenCode’s CEO, about what’s driving this wild growth, the Anthropic clampdown that inadvertently fueled it, and the almost 20 year founder journey that led him here.
00:44 — OpenCode’s Explosive Growth
01:16 — 20x Growth, 13M Users, and 7 Trillion Tokens
03:39 — The Anthropic Controversy That Changed Everything
05:43 — Bringing AI Coding Agents to the World
06:39 — When Open Source Models Became Good Enough
08:56 — What Millions of Developers Are Actually Using
13:31 — Why OpenCode Is Huge Outside the US
15:27 — Why Fortune 500 Companies Choose OpenCode
16:36 — The Economics of AI Tokens
20:02 — How Enterprises Are Using Coding Agents
22:58 — AI’s New Unit Economics
24:56 — Why Model Choice Matters
29:55 — The Product Decisions Behind OpenCode
34:21 — A 16-Year Overnight Success
41:16 — Why Jay Never Gave Up

I was at YC AI SUS last year.
Now I’m back, this time in @ycombinator with @insforge
If you’re going to AI SUS and want to maximize your chances of getting into YC, here are a few things I’d do differently:
1. Network with your peers, but choose wisely.
There are tons of afterparties and side events. You can’t go to everything. Spend time with founders you can actually learn from, build with, and stay connected with after the event. You can meet a lot of yc founders along the way 👀
2. Talk to YC partners.
This is one of the rare chances to get direct advice from YC partners. Use it. Last year, I pitched our idea to @aaron_epstein and got his feedback. Don’t be afraid to walk up, introduce yourself, and talk about what you’re building.
3. Leave with concrete action items.
This isn’t university. Don’t just sit through lectures, take notes, and go home. Build something. Talk to users. Try products from YC startups. Figure out what you can execute on immediately.
The biggest value of AI SUS isn’t what happens on stage.
It’s what you do with the people, feedback, and opportunities around you.

I asked @GarrettLord: "A young person watching this wants your life and what you've accomplished. What's the one thing you'd tell them this journey requires?"
His answer was 💥!
Nakul Mandan: "If you just compound and stack day after day after day and give it your best, great things will happen. That's been the story of my life and the story of this business."
@GarrettLord. Founder & CEO of @joinHandshake.
At a Christmas party in December 2024, Garrett met a

I’m a daily user and so happy for this team. Epic execution and they’re just getting started.
They process about the same volume of tokens as openrouter, which for context is rumored to be under talks for a $10B acquisition, and have 20x’d in a few months.
Y Combinator: Since the start of the year, @OpenCode (YC W21) — an open source alternative to Claude Code and Codex that works with any model — exploded to 4.6 million weekly active users, 13 million monthly actives, and roughly $40M in annualized revenue.
In this episode of The Lightcone,

Distillation is a research technique/algo. It's not inherently good or bad. If you want to do this kind of research, I'll fund great researchers who want to do this with $1M to do experiments that will make it even more sample efficient.
Then, we will openly release to the world - DM
Suhail: There is going to be 10x more distillation research with the threat of open weight model restrictions. It will Streisand effect the whole community into making it even more data / compute efficient. Just to troll the government. Silly game to play.

OpenCode now processes 7 trillion tokens per day. For context, that's more than all of Open Router. New Lightcone ep with @jayatvSST on how they got there.
Y Combinator: Since the start of the year, @OpenCode (YC W21) — an open source alternative to Claude Code and Codex that works with any model — exploded to 4.6 million weekly active users, 13 million monthly actives, and roughly $40M in annualized revenue.
In this episode of The Lightcone,

As Trump returns to tariffs, it's worth checking out what we've gotten from the trade war so far.

OpenCode is the first startup I’ve worked with where enterprises are pushing them to move faster through procurement so they can start using it. Customers want choice and that’s driven incredible growth for OpenCode this year.
Y Combinator: Since the start of the year, @OpenCode (YC W21) — an open source alternative to Claude Code and Codex that works with any model — exploded to 4.6 million weekly active users, 13 million monthly actives, and roughly $40M in annualized revenue.
In this episode of The Lightcone,

So much of being a founder is overcoming fear.
Fear of launching
Fear of talking to users
Fear of rejection
Fear of ambiguity
Fear of charging money
Fear of looking dumb
Fear of doing the hard thing
What are you avoiding because of fear? Don't let it control your company.

Spatial intelligence is the next frontier for AI.
During its Q2 earnings remarks, Google highlighted growing demand for AI infrastructure from robotics and spatial intelligence companies, including World Labs. ↓

Learning causality from internet videos in latent space first, and then using RL to teach the foundation model how to act.
This approach is 30× cheaper than Gemini 3.1 Flash on pretraining and achieves a better result. JEPA is all you need!
However, actions are still learned during post-training. Figuring out how to automatically learn actions without action labels is still very important!
Induction Labs: We’re introducing imagination models: a new foundation model architecture that unlocks learning from internet-scale video.
Our first imagination model, Photon-1, learned to use a computer by watching 18 years of screen recording video without action labels.

A big part of this is due to compression at the top end of the score distribution. If the SAT was considerably harder you’ll end up stratifying students better.
Though I don’t mind using various other metrics like high school debate / science fairs / competitions as reasonably proxies for what that would be
Deedy: It blows my mind that Northeastern is now a more selective school than Northwestern.
US college reputations have changed a ton in the last 15yrs.
— Vanderbilt / UChicago are now squarely elite: no different from MIT / Princeton in terms of SAT scores and selectivity.
— The SAT

Re I still remember the days when it was basically us & a few other people pushing this topic (on the top of my head @allen_ai @ylecun @jeremyphoward @soumithchintala @BlancheMinerva @fchollet @GuillaumeLample,…). We’ve gone a long way for it to be the ceo of the most valuable company in the world’s first tweet!

I previously reported that the growth rate and composition of Trump's wealth gain in office are unprecedented in recent history.
Upon request, here is the data on the relative size of Trump's presidential windfall. The extent of his self-enrichment is even more striking.

Breaking News: Federal officials acknowledged in court documents that they canceled grants to states that didn’t vote for President Trump in 2024. https://nyti.ms/4pIbybQ

Open source is a positive and important force for both empowering people and preventing centralization. Proud to support this.
Satya Nadella: Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. https://www.microsoft.com/en-us/corporate-responsibility/topics/open-weight/

Thanks for having me on The Lightcone podcast @YCombinator! Fun conversation, excited for it to come out! 🤓🚀
@garrytan @snowmaker @sdianahu @harjtaggar

Live now: our entire AI x Evals Track!
https://www.youtube.com/watch?v=q2JrUKBMf0w&list=PLJ7eF79yCUHc
- @aparnadhinak, CPO @arizeai
- Jason Lopatecki, CEO @arizeai
- @lukaspet, Co-Founder, Andon Labs
- @_philschmid, Staff Engineer @GoogleDeepMind
- @maorbril, Character[.]ai
- @ryfeus, Snorkel AI
- @guptasoumya12 + @jai_chopra, Uber
- @daverevere + Akele Reed, SonderMind
- @DanielJBump + Preetika Bhateja, Google
- Alex Shaw + Ryan Marten, Laude Institute

This is like the Delta Force of San Francisco nimbys
Laura Waxmann: Hundreds of #SanFrancisco Marina residents turned out this evening for townhall on controversial @Safeway housing plan, which would place what for many is an uncomfortably tall, 22 & 18 story project at the low-slung neighborhood’s waterfront.
@sfchronicle

I thought I was done building startups until I met @EmilyColman.
After YC W10, raising ~$100M, and building a $50M FDA-regulated manufacturing business, I wasn't planning to come back.
Then yesterday happened.
The FDA's peptide committee voted 'yes' to legalize peptides.
That's why I'm back in @ycombinator YC S26 launching @Ronan_RX out of stealth today.
Most of the news got It wrong. Here's what's actually happening.

Preparation in progress for the San Francisco Open Weights mini-march tomorrow. Hope all these big tech CEOs won’t show up otherwise I might end up in jail (haven’t had time to ask for a permit 😅)

The proposed Safeway housing development would probably yield more property taxes to fund schools, public safety, and affordable housing than the sum total of property taxes paid by everyone in that room.
Laura Waxmann: Hundreds of #SanFrancisco Marina residents turned out this evening for townhall on controversial @Safeway housing plan, which would place what for many is an uncomfortably tall, 22 & 18 story project at the low-slung neighborhood’s waterfront.
@sfchronicle

Scape became my primary email client for the last couple of months. What they built is inevitable and the new bar for how email clients should work. ✉️
Melvin Hagberg: We’ve raised $3.2M and spent 18 months building Scape.
Scape is the first email inbox that works for you:

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.
Laura Waxmann: Hundreds of #SanFrancisco Marina residents turned out this evening for townhall on controversial @Safeway housing plan, which would place what for many is an uncomfortably tall, 22 & 18 story project at the low-slung neighborhood’s waterfront.
@sfchronicle