Dan McAteer
Browser Harness:
Self-healing harness that enables LLMs to complete any task.
With all this talk about "intelligence per token"... well...
Matt Shumer: We need a new term like "Intelligence per second" to capture the magic of some of these newer models
Truly an honor and blessing to host @demishassabis at YC today 🙏
Dimitri Dadiomov
It might feel like we’re living in the steep end of the J-curve, but we ain’t see nothing yet. These two are doing as much as anyone to steepen the curve and bend the future closer - fascinating chat today with @garrytan and @demishassabis at @ycombinator
Bruno Koba
Demis Hassabis at YC today:
"We're only one or two technical breakthroughs away from AGI. But all the other parts are already in place."
Xchat is a banger
Only a matter of time before Cursor becomes Xcode amirite? 🙂
Teknium 🪽
I literally run 12 hermes agent instances every day in parallel to build Hermes Agent, and its now a top 100 GitHub repositories of all time. Agents do bring value and do create substantive software and work.
David Cramer: Everyone is slowly coming to this realization, and I assure you, no one is running multitudes of agents overnight. No one that is doing anything of substance at least.
There _are_ people pretending to be scientists, or fully caught up in their drug infused AI overdose, that
Demis Hassabis
Thanks for inviting me @garrytan, was awesome to chat and loved the inspirational space! Great to see so many startups building with @googlegemma models!
Garry Tan: Truly an honor and blessing to host @demishassabis at YC today 🙏
Peter Yang
Asked GPT 5.5 and Codex to make Star Fox. This is after 15 min of prompting 🔥
Peter Yang: Should I ask Codex to make Star Fox next
Davit
"What would the developer experience even be like for someone using a continual learning model? Any idea how you’d actually steer it?" @garrytan asked, @demishassabis answered.
"Not having continual learning right now is one of the things holding agents back from doing full end-to-end tasks.
They’re really useful for aspects of tasks today. You can compose them together and do some really cool things, but they don’t adapt well to the specific context you’re operating in.
That’s the missing piece for them to become truly fire-and-forget agents that can just figure it out themselves.
They need to be able to learn about the specific context you put them in. So I think we have to crack continual learning to reach full general intelligence."
Davit: Demis Hassabis and Garry Tan on the importance of context engineering and memory for continual learning!
Change the party to murder all capitalists (Piker) and destroy the schools (Bowman), brought to you by a tech centimillionaire who is using his money to buy an election (Chakrabarti)
No thanks!
the openai team ships
Kane 謝凱堯
The Insurance Commissioner is a down ballot election but has a lot of impact on cost of living.
Jane Kim is a horrible candidate from the Bad Days of SF’s recent past. She’s trying to use this niche election to re-enter politics.
Leave her vote blank. Tell your friends.
Jane Kim 金貞妍: The numbers just dropped and we have great news.. it's all because of you. THANK YOU!
inside me there are two wolves
John 🔱 Li: @aiDotEngineer @mattpocockuk 🤭
Paul Graham
Hamming's talk is so important that I reproduced it on my site. It's one of the only things on my site written by someone else.
https://paulgraham.com/hamming.html
Ihtesham Ali: A mathematician who shared an office with Claude Shannon at Bell Labs gave one lecture in 1986 that explains why some people win Nobel Prizes and other equally smart people spend their whole lives doing forgettable work.
His name was Richard Hamming. He won the Turing Award. He
Paul Graham
Hamming's talk is so important that I reproduced it on my site. It's one of the only things on my site written by someone else.
https://paulgraham.com/hamming.html
Ihtesham Ali: A mathematician who shared an office with Claude Shannon at Bell Labs gave one lecture in 1986 that explains why some people win Nobel Prizes and other equally smart people spend their whole lives doing forgettable work.
His name was Richard Hamming. He won the Turing Award. He
Re fun to think about what the pm thinks vs what the engineer thinks in this scenario
Ananyo Bhattacharya
23 years old with no advanced mathematics training solves Erdős problem with ChatGPT Pro. "What’s beginning to emerge is that the problem was maybe easier than expected, and it was like there was some kind of mental block.”-Terence Tao https://www.scientificamerican.com/article/amateur-armed-with-chatgpt-vibe-maths-a-60-year-old-problem/
Here's YC's official advice about being truthful and precise about what is pilot, bookings, revenue and recurring revenue.
Founders, particularly first time founders, need to sear this into their brains. Don't mistake one tier for another. Be precise, and always be truthful.
Science needs open source and open data now more than ever
Crémieux: Science is about to get absolutely nuked.
Unless we get extremely strict about providing and opening up code and data and documenting lab experiments rigorously, a torrent of credible-looking but fraudulent papers is upon us.
Daeshawn
Garry Tan: Here's YC's official advice about being truthful and precise about what is pilot, bookings, revenue and recurring revenue.
Founders, particularly first time founders, need to sear this into their brains. Don't mistake one tier for another. Be precise, and always be truthful.
My next guest is making $1M+ a month (!) from 5 AI products that he built as a solo founder.
Tomorrow, Tibo will break down:
→ How to validate and kill ideas fast
→ Why SEO is still his top channel in the AI era
→ The pricing sweet spot for AI products
Tibo also shared how 9 of his products failed before one hit $600K/month.
📌 Subscribe to get our full interview tmr: https://www.youtube.com/@peteryangyt?sub_confirmation=1
Peter Yang
My next guest is making $1M+ a month (!) from 5 AI products that he built as a solo founder.
Tomorrow, Tibo will break down:
→ How to validate and kill ideas fast
→ Why SEO is still his top channel in the AI era
→ The pricing sweet spot for AI products
Tibo also shared how 9 of his products failed before one hit $600K/month.
📌 Subscribe to get our full interview tmr: https://www.youtube.com/@peteryangyt?sub_confirmation=1
Erik Bernhardsson
Humble request to techies to stfu about AI mass unemployment and start to talk about using GPUs to cure cancer and find new materials and all the other amazing opportunities
Polymarket: JUST IN: An AI data center moratorium is now projected to pass this year as protests intensify nationwide.
85% chance.
GPT-5.5 raises the ceiling of ambition for what you can do with AI:
Simon Smith: Riley's recent tests and various posts about GPT-5.5 have revealed something about model progress. Models are already so good, you need to raise your ambitions or you won't realize just how good they've become. If you don't raise your ambitions, you'll think they've stagnated.
what are you building with codex?
Paul Solt: What app are you making this weekend with GPT 5.5 and Codex?
5.5 is so earnest
"little engine that could" energy
GPT Image 2 is great for learning
OscarAI: GPT Image 2 is also great for summarizing books or scientific essays through highly visual, detailed infographics.
Here I asked it for an infographic on On the Origin of Species by Charles Darwin.
Internet Archive
When MTV News shut down, it felt like decades of culture vanished overnight 🕳️
But over 470,000 pages were already preserved.
That history didn’t disappear.
It was archived.
📚 Read VANISHING CULTURE to see why it matters
📖 Download & read: https://archive.org/details/vanishing-culture-2026
🛒 Purchase in print: https://www.betterworldbooks.com/product/detail/vanishing-culture-a-report-on-our-fragile-cultural-record-9798995425014/new
#VanishingCulture #DigitalMemory #InternetArchive #MTVNews #BookTwitter
Y Combinator
Replit is the leading no-code app builder for consumers and enterprise, letting anyone with an idea build real, deployed software using natural language. The company just raised a $400 million Series D at a $9 billion valuation.
In this episode of Founder Firesides, co-founder and CEO Amjad Masad (@amasad) sat down with YC's Andrew Miklas (@amiklas) to talk about @Replit's 10-year journey from browser IDE to vibe coding platform, why the people getting the most value aren't traditional devs but founders and domain experts closest to the problem, and what Agent 4 unlocks with parallel agents, built-in design, and the ability to run your entire company on Replit.
0:28 - Anyone Can Build Software
2:14 - The Rise of AI-Native Builders
4:52 - Not Just Developers Anymore
7:18 - What People Are Actually Building
10:36 - How Replit Is Spreading Everywhere
14:02 - What You Can Build (and What You Can’t)
19:22 - YC, Growth, and Early Lessons
23:18 - From Vibe Coding to Autonomous Agents
29:44 - The Future: Everyone Becomes a Builder
36:12 - What Skills Matter Now
ChatGPT is just a less useful version of Codex that can't build and ship stuff.
Same for regular Claude vs. Claude Code.
It's all just prompting the AI at the end of the day.
Making stuff is fun
Liz4SF
if you haven't heard, YC launched a zine machine "where san francisco goes, so goes the world" w/ determination and deliberation that cant be ignored
https://thevoicesf.org/y-combinator-launches-its-zine-machine
Matt Carey
My talk at AI Engineer “Every API is a Tool for Agents” is out on YouTube
https://youtu.be/YBYUvGOuotE
Thanks to @swyx and the @aiDotEngineer crew for having me :)
I’m having my OpenClaw read my notes and writings from my time as an undergrad at Stanford and this was a particularly unusual find
David Mlcoch
I'm glad @ycombinator is advising founders to be truthful about their revenue numbers.
I remember our first office hours with @harjtaggar who I believe gave us a demo day goal of $10k MRR, a few paying customers who love us and more importantly figuring out who exactly our customers are. (Getting revenue is easier than that:)
I felt $10k MRR is nothing, I saw Twitter filled with exponentials of $$$ in 5 days. I felt YC was lowering our ambitions. Only later when we reached $10k MRR, I realized how hard the actual 0 -> $10k MRR in 3 months is and that it's a great achievement.
Garry Tan: Here's YC's official advice about being truthful and precise about what is pilot, bookings, revenue and recurring revenue.
Founders, particularly first time founders, need to sear this into their brains. Don't mistake one tier for another. Be precise, and always be truthful.
GStack v1.13 shipped - simple feature, I'm spending a lot more time using GPT-5.5 with Conductor and so I realized I sometimes wanted to get Opus 4.7's take on a change before I kept going. Cross-modal synthesis is something I find myself relying on more and more.
People on X:
“Looks fun, do a barrel roll!”
People on Threads:
Peter Yang: Asked GPT 5.5 and Codex to make Star Fox. This is after 15 min of prompting 🔥
The Philadelphia Inquirer
OPINION: "Trump’s mismanagement has resulted in death, destruction, wasted tax dollars, and a huge loss in respect for the United States — all to accomplish nothing." https://ebx.sh/V5TJjQ
GBrain v0.21 now adds code graph support to our existing retrieval for non-code: graph, vector, hybrid RRF, and of course grep
I'm working on revisions to GStack to support this code graph AND incorporate memory and dreaming from your code sessions into your GBrain instance
Adish Jain ☕️
i just don't understand this take.
most people going thru @ycombinator are first-time technical founders that are tinkering and building.
it's not like you become a founder and then you're suddenly instilled with an understanding of the differences between LOIs, cARR, annualized revenue run rate, ARR, etc.
most of YC is spent building something people want and most of being a founder is about learning as you go.
and in a world where founders are incentivized to inflate their revenue to look like they're keeping up with the speed of AI and ever-moving goal post of "success" to unlock funding, laying things out in simple terms is not shocking. it is necessary.
Anthony Lombari: ngl its kind of shocking we need to lay this out like this?
This should already be known to founders.
btw we are cooking something with @hhua_
(not final yet but keep calendar open after ICML in Seoul)
DeepSeek: 🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params.
Chamath Palihapitiya
On page twenty-six of “The Billionaire Tax” proposal in California, it explains how the state legislature can convert from a Billionaire Tax to an Everyone Tax without voter approval.
They can also adjust the tax to be a yearly tax, not just one time…again, without your approval.
Intelligence test for you: if this was meant to just target Billionaires, why did they write this in?
Chamath Palihapitiya: The Billionaire Tax is actually an Everyone Tax.
The Billionaire Tax is a new tax proposal written by four professors who don't believe in the American dream. Some of them aren’t even American…go figure.
Despite its name, it applies to every California resident who currently
I often have to get Codex to fix my OpenClaw set up 🥲
بدر الحناكي
سويت لعبة تأثير ستروب مرة ثانية
Vox
gbrain 0.21 lets your OpenClaw / Hermes read code through a code graph.
→ sharper debug: ask about one function, get the body + who calls it + what it calls + parent class scope. no more isolated chunks
→ see what gets affected before you change a line
→ PR review can ask which call sites a change actually touches
→ sharper code search
→ language coverage works for us: TS/TSX/JS/Python fully covered
plain version: help your agents actually understand code structure now.
Garry Tan: GBrain v0.21 now adds code graph support to our existing retrieval for non-code: graph, vector, hybrid RRF, and of course grep
I'm working on revisions to GStack to support this code graph AND incorporate memory and dreaming from your code sessions into your GBrain instance
GPT Image 2 for learning about endangered animals
Harboris: GPT Image 2 on ChatGPT
Prompt
Create a visually rich infographic about an endangered animal. Start by finding one online, research its habitat, diet, and unique traits. Present information through annotated visuals and structured callouts, not generic sections. Style it like a
Jason Freedman
This is the same advice I heard 10+ years ago. YC hasn’t changed much in this type of guidance.
When founders mix one or another up…(usually because they’re young and inexperienced) it takes like 90 seconds of questioning to get to the accurate perspective of the revenue progress. They often fix the slide live on our call with them.
Is everyone else really struggling here? Or do people just enjoy the rage baiting at YC?
Also FWIW, revenue metrics (of all kinds) are not very predictive at this early moment. John and Patrick were in my batch and had no revenue (nor CARR/pipeline/ARR) for dev/payments (Stripe). A lot of startups were ‘farther along’
Garry Tan: Here's YC's official advice about being truthful and precise about what is pilot, bookings, revenue and recurring revenue.
Founders, particularly first time founders, need to sear this into their brains. Don't mistake one tier for another. Be precise, and always be truthful.
how can they write code so fast?!
henrique cunha: codex app is trending to be the best software i've ever used
ridiculous how fast it got so good
GPT-5.5 for the enterprise:
Databricks: .@OpenAI GPT-5.5 is now available on Databricks, with Codex coding workflows and model inference fully governed through Unity AI Gateway.
With GPT-5.5 on Databricks, you can:
- Power coding workflows with Codex or other coding agents
- Build custom agents grounded in enterprise
Kevin Smith
Such a fun @snipd_app user meetup in SF today :)
Today, we had a Snipd user join us who remixes parts of podcasts into music tracks 🤯 -> new feature idea? 🤔
Thx @swyx & @cognition for hosting us!
oh, *that’s* what tenet was about
Umesh: ChatGPT Images 2.0 explains “Tenet” in a simple way!
oh, *that’s* what tenet was about
Umesh: ChatGPT Images 2.0 explains “Tenet” in a simple way!
Gustaf Alströmer
Shortage of housing is the source of many of the most critical problems in the San Francisco and the bay area. Scott Weiner have done more than anyone to fix it. I'm voting for him on June 2nd. @Scott_Wiener
SB 35 (2017) — Streamlined housing approvals in cities failing to meet state housing goals, his signature housing law.
SB 9 (2021) — Upzoned most of California to allow up to 4 units per single-family lot.
SB 79 (2025) — Legalized mid-rise apartments around major transit stops, a decade-long fight finally won.
Kim-Mai Cutler: There’s only one candidate in the SF Congressional race with the receipts to show they’ve been working on housing for 10+ years.
There are finally towers being proposed near me in the last month.
Would have been impossible without him.
GPT Image 2 for changing the style of any photo of yourself or your family
Linus ✦ Ekenstam: You can try this:
Turn any photo into a beautiful woodcut/linocut style, GPT Image-2 does a great job with details, expressions.
Perfect thing for the profile picture or the family photo. Or why not a gift?
Try for yourself, full prompt below ⤵️
GPT Image 2 for changing the style of any photo of yourself or your family
Linus ✦ Ekenstam: You can try this:
Turn any photo into a beautiful woodcut/linocut style, GPT Image-2 does a great job with details, expressions.
Perfect thing for the profile picture or the family photo. Or why not a gift?
Try for yourself, full prompt below ⤵️
GPT Image 2 for reimagining damaged photos:
WasifAI: Age-worn, damaged images can now turn into 4k with just one prompt using Chatgpt. & Its Free 💸
Made on Chatgpt, GPT-2
Prompt: 👇🏻
GPT Image 2 for reimagining damaged photos:
WasifAI: Age-worn, damaged images can now turn into 4k with just one prompt using Chatgpt. & Its Free 💸
Made on Chatgpt, GPT-2
Prompt: 👇🏻
x.com