Great to see a GBrain x ActiveGraph crossover here
Great retrieval pairs well with almost anything you want to do with agents
Apprentice AI: Gbrain is what an agent *knows* — a durable markdown/git knowledge substrate with hybrid search, typed links, facts, timelines, and a real ingestion pipeline.
@ActiveGraphAI is what an agent *did and why* — an event-sourced runtime with replay, fork/diff, and full causal
Saikat Chakrabarti said that a single member of Congress from San Francisco is "useless."
This type of statement should immediately disqualify him from office. He doesn't want to represent the people.
He just wants to buy personal power and clout.
He is no public servant.
David Klindt
What does JEPA actually learn? We can finally prove it 🌍
So excited to share our theory of identifiable World Models: LeJEPA recovers the latent variables of the world.
Plan in the learned World Model as if it were real, same shortest path.
📄: http://klindtlab.github.io/lejepa-identifiability
David Klindt
Re Identifiability is what it means to learn the right latent space, the Gaussian is why LeJEPA works, and it's what makes planning transfer.
I think identifiability is the right definition of what it means to learn a World Model 🌍
The natural next step for the theory: add action conditioning, like in LeWorldModel (@lucasmaes_ et al.),
w/ @randall_balestr @ylecun 🔬🤖
Paper: https://arxiv.org/abs/2605.26379
Code: https://github.com/klindtlab/lejepa-identifiability
Feedback is a gift. Critical feedback doubly so.
Next.js Night · AMS 🇳🇱 · June 11
Learn what’s next + meet the team + share feedback
http://luma.com/34nqdfc3
Madhav Singhal
latent space pod is still very underrated, especially on the applied side of things
glad that they have kept it very high signal
one of the few pods i listen to while working out and I always find myself re-looking up some reference/info that i found on the pod for work
otterboyinfrisco🏳️🌈
It really is not hyperbole to say that Peskin and SF's "progressives" really genuinely came up with the worst housing policies on literal planet earth for 25 years+.
Genuinely don't know how SF gets out of this.
donkeyDdevelops: Crazy that SF rents are at all time highs and still nothing pencils there for ground up.
(actually maybe not that crazy)
🛫
Hermeus: Supersonic. Mach 1.21.
Quarterhorse Mk 2.1 is now the world’s first privately developed, unmanned supersonic jet and the fastest unmanned aircraft flying today.
This flight makes Hermeus the fastest company in aviation history to go from founding to supersonic flight - exactly
GStack is still getting better daily. Just shipped v1.47
Thanks @jayzalowitz for the PR idea
Extra useful for when I need to fix GBrain issues and need to create a precise spec about the bug directly out of the OpenClaw instance that hit the issue
Matt Lieberman
I signed it. The data are clear. Standardized tests are the best predictor of college success and the least biased indicator we have. Getting rid of them was incredibly foolish.
Steve McGuire: "Current admissions practices do not provide a sufficiently reliable check on mathematical readiness for STEM majors."
Over 280 University of California STEM faculty have signed an open letter calling on the Board of Regents to reinstate standardized testing in admissions:
Brad Gessler
Re @nbaschez @gbrainio @garrytan Yeah. We setup a gbrain for a team two weeks ago and it took 90 min, so we got it down to 90 seconds on http://gbrain.io
Still a lot of work in front of us, but if you’re up for a journey, we’re steadfastly building a sane managed team OpenClaw + gbrain + UI.
Honored to receive a medal from his Majesty King Abdullah II for Distinction on Jordan’s 80th Independence Day.
It’s been an incredibly journey building @Replit, starting from Jordan more than 15 years ago, and then helping accelerate the field of agentic AI globally.
codex is great for any kind of work done with a computer:
Brandon Gell: If you know one thing about every right now, it's that we're heavily Codex pilled. So we wrote a guide on how to use Codex for knowledge work as well as we do. You dont want to miss this one...
https://every.to/guides/codex-for-knowledge-work
Founders must stop trying to building 2010-era businesses with 2026-era technology.
Don't try to rebuild Foursquare or Yelp.
Don't try to recreate Basecamp by 37 Signals with $10/mo SaaS pricing.
Don't underprice! If it works it's worth a lot more.
Don't be tempted to become "Tech enabled PE" with revenue tricks.
The rules of tech changed with AI. Play the new game.
Y-3
AI-Literacy and Reading Between the Slop
https://yyy3.substack.com/p/ai-literacy-and-reading-between-the
NIK: > I’m now in the LeCun/Marcus camp on LLMs
> real programming agents will need world models
> not some RLVR shit
it’s over
Moses Xu
the mvp rule made sense when humans were the execution layer: cut scope, ship fast, iterate.
agents change which work is expensive.
systems aren't procrastination now, they're the only thing that scales.
Peter Yang: What used to feel like procrastination (building systems instead of the MVP) is now a prerequisite to ship effectively with AI agents.
My number 1 lesson from @ryancarson:
"We used to say just do the bare minimum to get the MVP out. Don't spend time on systems. It's literally
true but changing fast
Austen Allred: Codex remains underrated
Jason ✨👾SaaStr.Ai✨ Lemkin
New!! "Why This is The Best Time Ever To Be Building, Where AI Agents Are Going, and Much More with Replit"
https://youtu.be/QFZH6_3HdTo
gm
Track day.
Amjad Masad: Back in Jordan doing my favorite thing — drifting! First time in a pro drift car.
Aaron Horne
Try DemoGuards today. 14-day free trial, no CC required for trial. https://demoguards.com
Created on @Replit, as well as the animation and video
Alex Rives
Today we're announcing ESMFold2, an open scientific engine to power prediction, design, and discovery across protein biology.
The new model delivers state of the art performance on protein interactions, especially antibodies, a critical modality for therapeutics.
We have designed and validated miniprotein binders and single chain antibodies across five therapeutic targets that are important in cancer and immunology. We are seeing very high success rates, and affinities at levels consistent with therapeutic activity.
We’re also releasing an atlas of 6.8 billion proteins, and 1.1 billion predicted structures.
ESMFold2 is built on a state of the art language model that has been trained on billions of protein sequences.
A world model of protein biology emerges through language modeling.
We’ve used the techniques of mechanistic interpretability developed to understand large language models to understand the concepts ESM uses to represent proteins.
The model’s representation space has a compositional organization of features across scales, levels of complexity, and abstraction, that reflects and mirrors the understanding of protein biology developed through a century of empirical science.
This understanding emerges without prior knowledge, just from language modeling of protein sequences.
Language models are becoming a powerful substrate to understand and program biology.
The design of protein interactions is one of the most fundamental problems in biophysics, and has critical implications for the discovery of new medicines. A simple gradient based search with the model was able to discover high-affinity protein binders.
I'm excited by the potential this has to accelerate basic science and the understanding of proteins. And especially for the new avenues it opens up for therapeutic design and medicine.
Every time GitHub has an outage our team is paged. Incidents at Vercel get automatically filed by anomaly detection systems.
We just detected an outage 16 minutes before their status page changed. Deployments suddenly dipped and surged.
Despite all the chatter about coding AGI, the reality is that software infrastructure remains an extremely hard problem.
I have no doubt the GitHub team is highly competent, and there's no shortage of models and agents available to them. Don't forget this is the company that brought us Copilot, the first major breakthrough product in AI coding. Yet clearly the prompt "/goal scale GitHub, make everything extremely fast, make no mistakes" is not enough.
The hard parts of software remain very hard, especially under unprecedented demand, as more people join in on the fun of building new things.
nature
A newly released AI tool has generated an atlas of more than one billion predicted protein structures and billions more protein sequences.
https://go.nature.com/4fblM0Z
Anish Acharya
the big thing that's changed since 2010 is you can be dramatically more ambitious in the type of software you build and charge dramatically higher prices
the durable lesson from foursquare and yelp and many 2010 era startups is to build at the edge of new tech and new culture
Garry Tan: Founders must stop trying to building 2010-era businesses with 2026-era technology.
Don't try to rebuild Foursquare or Yelp.
Don't try to recreate Basecamp by 37 Signals with $10/mo SaaS pricing.
Don't underprice! If it works it's worth a lot more.
Don't be tempted to
Liam
http://x.com/i/article/2059578647653285888
Liam
$4 million and 18 months. Then $200 and 5 days. Same product, rebuilt by one man.
That gap is the proof of what @garrytan just named: process power, the one moat anyone can build for themselves.
Everyone is quoting the price. The price is the boring part.
What matters is what he did with the time the cheap building gave back: he built a system that builds, and kept it as one person. He showed the whole loop and barely anyone caught it.
I wrote the full breakdown, including the one test that tells you if you are running his loop or just renting a chat window.
Garry Tan: The reason why I release my X articles about AI agents (fat skill fat code thin harness) and GStack and GBrain is that we, yes you and I, can have *PROCESS POWER*, which is the one super powerful specific moat that anyone can create for themselves.
The agent helps you do it.
Randall Balestriero
It is refreshing to have some new arguments pointing at specific distributions (Gaussian) our embeddings must follow! Especially now that we know how to do scalable distribution matching in high dimension with LeJEPA/SIGReg! Another win for Abraham de Moivre and Friedrich Gauss!
David Klindt: What does JEPA actually learn? We can finally prove it 🌍
So excited to share our theory of identifiable World Models: LeJEPA recovers the latent variables of the world.
Plan in the learned World Model as if it were real, same shortest path.
📄: http://klindtlab.github.io/lejepa-identifiability
Chris Murphy 🟧
This Ebola outbreak has been a slow motion disaster - 900 cases and and now threatening the U.S..
We know how to stop outbreaks like this. But Trump chose not to stop it. He destroyed our global health team, deliberately exposing us.
1/ Here's the story of how this happened.
ABC News: The CDC sent an "urgent request" to its workforce to recruit personnel to help screen passengers coming from Central Africa for any potential signs of Ebola illness, according to an internal email to staff obtained by ABC News. https://abcnews.link/yji3ra3
Zeming Lin
Re ESMFold2 is a state of the art folding model. It's crazy good. The -Fast model does better on antibody-antigen complexes than AF3 with MSAs.
Zeming Lin
Re First of all, all of this is open sourced. Even the binder design protocol [coming soon], which has been rigorously experimentally validated.
Paper: https://biohub.ai/papers/esm_protein.pdf
🤗: https://huggingface.co/biohub
Atlas: https://biohub.ai/esm/protein/atlas
Website: https://biohub.ai/esm/protein
biohub
Proteins are the machinery of life. Scientists have cataloged billions of protein sequences—but their biology is still mostly unknown.
Today we're releasing a world model of protein biology: a scientific engine for prediction, design, and discovery that consists of ESMFold2, ESMC, and ESM Atlas. Together, they're helping to open up a new way for researchers to design proteins and speed up scientific discovery.
Our mission is to cure or prevent disease. To do that, we need to accelerate science. That's why we're releasing all three openly. https://bit.ly/3PGf1dk
Rebecca Bakels
Who will I see at @aiDotEngineer World's Fair? Getting so excited!
Jitesh Ghanchi
I can't believe i'm using the same AI coding setup as the ceo of @ycombinator to build my products
@garrytan's gstack is awesome.
no millions, no SF, no connections needed
golden era for solo builders fr
Peter Yang
I got tired of making PowerPoint slides so I built an AI skill to do it for me.
Here's my new tutorial on how to build a /slides skill that turns a rough outline into a beautiful HTML deck in minutes.
I walk through how to:
→ Use 12 slide formats and 3 templates
→ Add live charts and subtle animations
→ Get AI to screenshot each slide and fix layout issues itself
📌 Watch now: https://youtu.be/vbChRIIlSPE
Paul Graham
Gustaf (right) when he was in YC. You can see why our nickname for this startup was "the band".
The best pitch test: can you tell this at a bar to a friend?
Not the market size, not the TAM, not the deck. Just the story. Why you built it. What you saw that nobody else saw. What happened.
If you can't tell it naturally over drinks (or it would feel weird to your friend and they would think "this isn't you"), your idea or pitch or endeavor needs to bake some more.
The 9pm-at-a-bar test for your startup idea is a defining authenticity test. And without that, nobody will buy your product, let alone come work for you or invest.
Kevin Dahlgren 🥾 🥾
San Francisco: A Shoplifting Paradise for the homeless. I visited San Francisco to document the homelessness crisis and witnessed some of the most blatant shoplifting I’ve ever seen. Every store I entered, someone was actively stealing while security guards stood by, often
http://x.com/i/article/2059649519722307584
I get up in the morning to be a witness to greatness in YC founders.
Garry Tan: http://x.com/i/article/2059649519722307584
Nick Mehta
Re @garrytan's quote below around using AI to create new products and services is so spot on - and so hard to do.
5 Failure Patterns in New Product Introduction to Avoid in the AI Era:
Every established company has face-planted one or more times trying to get into a new market. Think of all of the failed product launches from even the most iconic companies like Google, Amazon and others.
In my experience, there are at least 5 common patterns:
1. No "Right to Win." This is the biggest. What do you bring to the table in the new market that helps you win? Amazon had no right to win with the "Fire" phone. Apple had a right to win with AirPods (due to iTunes and iPhone). AI opens up infinite market - so choose the part where you are the logical winner.
2. Inferior v1. So many large companies have the hubris to think they can enter a new category with a "good enough" product. In the olden days where transparency was limited and switching was hard, this worked (e.g., Oracle). Now, every new space is a new fight. So you need to have a great product from day 1. AI obviously helps a lot here in terms of iteration speed. But don't let your incumbency go to your head.
3. No Sales Channel. Another challenge (particularly for B2B enterprise businesses) is they have an inability to sell the new offering. Maybe it's too cheap and the sales force focuses on the core. Maybe it requires a product-led motion and your main business is sales-led. This is where using AI to give the builders control of the go-to-market is huge. Before, product people would wait months for marketing to build a separate PLG site or for campaigns to go out. Now, speed can be rapid.
4. No Portfolio. The truth is that new products are bets. Author Jim Collins famously advised "fire bullets, not cannonballs" - meaning try lots of small things first. The problem is that in the pre-AI world, it was expensive to test new products. Hence companies often did it serially and with a lot of resource. Now, parallel processing is possible.
5. "Limited Talent." There was a perception that larger organizations failed to launch new products due to a lack of talent. I think this is partially true. I believe much of the talent was suffocated under a mountain of "cross-functional alignment meetings" for the new business. These are soul-crushing and demotivate the people that you want to be fired up. Free them by giving them autonomy and using AI as a part of that.
What else would you add in terms of why incumbents fail in new markets and how AI can help?
Garry Tan: @karrisaarinen Use AI effectively to create new products and services that didn’t exist before that customers love
Y Combinator
Over the past year, we've been building our own internal agent infrastructure at YC: over 350 tools, self-improving skill loops, and a shared organizational brain that gets smarter overnight.
In this episode of the @LightconePod, we sat down with YC General Partner Pete @koomen to talk about how he led the effort from the ground up.
We cover how giving agents unrestricted access to one database was the key unlock, the self-improving skill loops that get smarter overnight, and why he thinks we've arrived at the personal computer moment for AI.
00:39 — YC's AI Stack
02:15 — The Finance Team Problem That Started It All
05:07 — SQL Access Changes Everything
07:20 — One Database to Rule Them All
09:14 — Jevons Paradox
10:07 — Denormalizing for Agents
12:15 — The Single-Player Era of Agents
14:16 — 350 Tools and a Shared Registry
16:24 — Skillify, DRY, and MECE Resolvers
18:23 — The Self-Improving Dream Cycle
20:26 — The Two-Sentence Pitch Skill
23:06 — How Super Intelligence Compounds
25:10 — Recording Everything as a Building Layer
27:10 — The Shared Organizational Brain
29:18 — Trust-Default Culture as a Requirement
30:44 — Raising the Floor for New Employees
32:35 — Horseless Carriages
34:24 — Why Chat Is the Best Interface for Agents
38:50 — Just-in-Time Software
40:49 — Centralizing vs. Decentralizing AI
43:32 — The Personal AI Revolution
Zeming Lin
Get a state of the art protein structure prediction in just 9 lines of code. No MSAs needed.
Zeming Lin: I'm so excited to show the world what we've been working on the for the past months!! I'm going to highlight some of the fun results from this paper that I find particularly exciting.
Codex for transcribing and answering questions about a meeting in real time:
Simon Smith: You can now transcribe meetings in real time using Codex and ask Codex questions about meetings as they're happening!
I updated my new Codex Meeting Recorder skill to use GPT Realtime Whisper. Tell Codex to use the skill, and it will start transcription and show it in the
New in the Claude Marketplace: @augmentcode, @boltdotnew, @coderabbitai, @hebbia, and @WeAreLegora.
Apply your existing Anthropic spend commitment toward their Claude-powered products.
Learn more: http://claude.com/platform/marketplace
Claude: Introducing the Claude Marketplace, a way for enterprises to simplify their procurement of AI tools.
Now in limited preview.
If you abhor business and prosperity it will go elsewhere
California and blue states need to stop voting by sound bite and start voting for prosperity again
Joshua Rauh: California's "anchor" firms are growing fast - but they're doing it in other states.
With @LincGerstner we looked at 23 large public companies with strong CA employment presence: Alphabet / Google, Amazon (incl. AWS), Apple, Applied Materials, AT&T, Broadcom, Charles Schwab,
Legalize standardized testing
Legalize merit
Neetu Arnold: University of California STEM professors want standardized tests back due to severe math deficiencies among students:
“We now observe preparation gaps so severe that instructors must reteach middle school mathematics”
“The current admissions metric, based primarily on GPA &
@johnloeber: There are several interesting things going on here.
(1) The UCs were long known for their high standards, in particular UC…
Harm reduction has become harm acceleration
Defund do-drugs-until-you-die policy and fund treatment and recovery
Amber Richmond: The people being handed needles on the street are somebody’s mom. Somebody’s dad. Somebody’s child.
I was one of them.
I had access to clean needles and still ended up needing two open heart surgeries.
If I walked up to a needle exchange right now with almost 6 months clean,
Pete Koomen
For more than a year we've been building tools at YC to make it easier to use agents at work. I enjoyed speaking publicly about this for the first time with the Lightcone crew!
Y Combinator: Over the past year, we've been building our own internal agent infrastructure at YC: over 350 tools, self-improving skill loops, and a shared organizational brain that gets smarter overnight.
In this episode of the @LightconePod, we sat down with YC General Partner Pete @koomen
AI should dramatically increase quality of life and individual freedoms for people around the world.
The OpenAI Foundation is making an initial $250M commitment to measurement, transition support, and new approaches to broadly shared prosperity.
http://openaifoundation.org/news/economic-futures-in-the-age-of-ai
Ankit Gupta
why join YC? My two favorite reasons:
(1) every YC partner is a former YC founder that has built/scaled a company
(2) YC itself is a cutting-edge technology company, built and led by the partners
we will always live in the future and be the earliest adopters of technology
Y Combinator: Over the past year, we've been building our own internal agent infrastructure at YC: over 350 tools, self-improving skill loops, and a shared organizational brain that gets smarter overnight.
In this episode of the @LightconePod, we sat down with YC General Partner Pete @koomen
Claude
Michele Catasta (@pirroh) is President and Head of AI @replit, the platform where anyone can build software in natural language.
At 16, he set out to make software open to everyone. Today, over 50 million people are building on Replit with Claude:
Justin Kan
Finally got agent pilled and set up Hermes agent on Opus, http://you.com for search, connected to whatsapp, told it to install Gstack on itself. Feels like I'm playing Productivity(TM): The Video Game. I'm addicted.
Eric Levitz
Trump could have presided over a very strong economy. All he had to do was...nothing: Inflation was trending downward, the AI boom was raising growth.
Instead, he engineered a global trade war and energy shock, while shrinking America's labor force. https://www.vox.com/politics/489397/inflation-prices-iran-war-tariffs-trump
The Deployments list was one of the earliest views I built on the (zeit) platform.
One of the earliest 'acid tests' of Next.js was whether it could handle the dynamism of an app and the speed of a website, in one framework.
Redesigning a view that's so heavily depended on is no small feat. While I didn't build this one, it was fun to be flagged-in to opine on the many variations the team tested to get this out 😁
Vercel Developers: The redesigned deployments list is here.
See more deployments at once with a denser layout and an improved mobile experience.
https://vercel.com/changelog/redesigned-deployments-list
insanely good company to keep
Latent.Space: 🆕Railway's Agent-Native Cloud: 3M users, 100K signups/week, $200K+ coding agent spend, production forks, & the death of PRs https://latent.space/p/railway
@Railway founder @JustJake explains why AI agents need a new cloud, how Railway moved the vast majority of workloads onto its own
Georgia Channing
today was a massive day for protein engineering.
esmfold2 dropped—next gen of the esm series, fully open on @huggingscience. 1.1 billion predicted structures, 6.8 billion sequences. 800m more entries than the alphafold db, and reportedly edging out alphafold3 on protein complexes, including antibody–antigen binding.
alongside it: the new esm atlas. a huge expansion of known protein space, heavy on metagenomic sequences from soil, ocean, and the parts of biology that have been least characterised (until now!!)
and if that weren't enough, litefold dropped the fineweb of proteins, so every major protein database (pdb included) aggregated, cleaned, and made plug-and-play in one place.
these are the releases that push the whole field forward, and the pace of open science right now is almost motion-sickness inducing
all of it on http://huggingscience.co (and ofc @huggingface)
Last week at I/O, we got to hear from the writers, designers, coders, and filmmakers behind products like @FlowbyGoogle, Project Genie, and @GoogleFlowMusic.
A key theme: “See technology as a blank canvas… it's really the human imagination that shapes the technology into new stories and new ways of telling stories” as Matthew Carey puts it. Even as tools evolve at lightning speed, human collaboration remains the absolute center of the creative process.
@KhyatiTrehan and @kaloyaan, pictured below, also shared how they blend their traditional creative design knowledge with vibe-coding to create the vision for Google Flow Tools. As Khyati said: "If [the right tool] doesn't exist, maybe I can just make it." Kaloyan sees this newfound accessibility as an explosion of new creative possibilities and ideas we have yet to see.
Go explore products like Google Flow, Project Genie, and Google Flow Music at http://labs.google.
Kyunghyun Cho
join us at NYU Global AI Frontier Lab!
@c10labs , @nyuniversity and @NYCEDC invite you to an afternoon bridging academia and industry. Student researchers and early-stage startup founders will deliver lightning presentations on work at the frontier of AI, biotech, and hard tech — followed by a panel with investors and academics on what it actually takes to nurture the next generation of innovators.
rsvp link below!
🔥🔥🔥
Sergei: 4 years ago, I moved to New York to help start the Replit office in Brooklyn.
Today, I walked by our billboards in front of Madison Square Garden. Proud!
Use Runway in Replit
Runway: Introducing Runway MCP. Now you can connect Runway directly into Claude, ChatGPT, Cursor, Replit and more.
Generate polished images and videos with state-of-the-art models, like Gen-4.5, Seedance 2.0, GPT Images 2.0, Kling and more. Right from where you're already working.
biohub
A closer look at what we released today 🧵
ESMC is a language model trained on billions of protein sequences spanning the full diversity of life. Trained across 2.8 billion sequences, the model is exposed to the full distributional structure of evolutionary selection, and in learning to predict it, develops internal representations of protein biology. https://bit.ly/3PGf1dk
AYi
看完 Garry Tan 那篇关于 YC 的长帖， 大半夜的失眠了，想了好久。
心理学家 Alice Miller 一辈子在追一个问题， 为什么同样在恶劣家庭里长大， 有的孩子成了正常的大人， 有的孩子一辈子困在创伤里。
她最后给出的答案有点反直觉， 跟韧性、性格、治疗都没关系。
关键是身边至少有一个人， 愿意在那个孩子说这一切不对劲的时候， 看着他的眼睛说一句， 我看见了，这是真的，你没疯。
Miller 把这种人叫启蒙见证者。
Garry Tan 把这个概念搬到了创业上， 说 YC 真正的核心其实是那张餐桌上的某个人， 能够轻松说出一句， 我见过和你一模一样的开局，他后来做成了。
钱和那张沃顿名片只是表层， 那张餐桌后面累积的 6000 家公司的样本， 才是真正的护城河。
说白了，每个创业者刚开始都活在一个奇怪的处境里。
你父母觉得你疯了， 朋友圈觉得你在玩， 前同事一边祝你成功一边等着看笑话。
我觉得这倒不是他们坏， 主要是他们这辈子没见过两个人从公寓做出十亿美金公司的样子， 没见过的事，对他们来说就是不存在的。
你就像一个人在地下室挖矿， 所有人在地面上喊你回来， 但你听见了岩石另一边的声音。
YC 那个伙伴的作用其实简单， 他只做一件事， 把头探下来看一眼你挖的洞， 然后告诉你这里六年前另一个人也挖过，他挖通了。
仅仅这一句话， 你心里那种自己是不是真的疯了的疑问， 会立刻塌下来。
我盯着这段看了很久， 因为它讲的早就不止是 YC 了。
每个做内容、做产品、走非主流路径的人都缺这个， 缺一个见过相同地质层的人在旁边点一下头。
朋友的善意没法替代见证者， 善意的关心有时候反而是更精致的怀疑。
真正能让你穿过那段最黑的隧道的， 是某个看过 6000 次同样故事的人， 在你最虚的那个瞬间， 轻轻说一句， 我见过，你没问题。
所以说，有时候改变一个人命运的，根本不是被说服，是被看见啊。
@garrytan thanks Garry！Your writing really cheered me up.
Garry Tan: http://x.com/i/article/2059649519722307584
Replit supported more than 3,000 Saudi students to build their apps — many went on to become successful projects!
نون التعليمية: ضمن الشراكة بين نون التعليمية ووزارة الاتصالات وتقنية المعلومات، @McitGovSa وبالتعاون مع @Replit ، أُطلق “هاكاثون رواد المستقبل” لتمكين طلاب الثانوية في الذكاء الاصطناعي التوليدي، بمشاركة +70 طالبًا وطالبة.
امتدادًا لنجاحات مكّنت +3,000 طالب من تطوير +10,000 مشروع تقني ضمن
Noah Smith 🐇🇺🇸🇺🇦🇹🇼
FUCK YEAHHHHHH
ANTHROPIC IS NO LONGER DOOMING ABOUT JOBS!!!!! ♥️♥️♥️
cognition is now the largest independent agent lab in the world. take the 200% utilization that everyone is hitting from this chart and run out the sales growth from this, i encourage you to go thru the exercise if you are new to investing
a lot of you have read my cog initation post, but when i talk to people they do not sufficiently appreciate that this is the perfect storm of all the trends:
- first* koding agent
- first* cloud dev infra
- best reviewed code review/security guy
- first* llm wiki knowledge base
- s-tier GTM
- most IOI golds
- somehow also cracked at Smash Bros and Poker
agent lab gets you:
- long model diversity
- long reasoning/toolcalling models
- long harnesses
- long domain specific RLFT
- long coding data/expertise/evals
- long full agentic SDLC
- chief partner to CIOs in combating tokenmaxxing slop from humans and from agents
- SOTA foodbench from house chefs
and this is why you can see this is Peter Thiel's biggest AI bet
*i dont like "first" as an adjective, but "first + 2 years of serious scaling" really serves as a shorthand for "most enterprise battle tested" - see customer list in screenshot - literally 10s of thousands of devs, 10s of thousands of repos, PER CUSTOMER is the kind of weight I put on the simple word "first"
Cognition: 1/ We’ve raised over $1B at a $26B valuation, led by @Lux_Capital, @generalcatalyst, and @8vc.
Our enterprise usage has grown >10x since the start of this year, and our run-rate revenue grew to $492 M.
We launched Devin two years ago as the first AI software engineer. Since
Mayor Matt Mahan
UC STEM faculty are absolutely right, and this change is long overdue.
We’re doing our kids a disservice and putting at risk the reputation of the world’s best public university system. As Governor, I’ll bring back SAT/ACT scores as part of the admissions criteria for STEM degree applicants.
Steve McGuire: "Current admissions practices do not provide a sufficiently reliable check on mathematical readiness for STEM majors."
Over 280 University of California STEM faculty have signed an open letter calling on the Board of Regents to reinstate standardized testing in admissions:
Michele’s been applying AI on code since before LLMs existed, which is why frontier labs enjoy collaborating with our team.
It’s been super fun working with Anthropic to constantly find the absolute edge of what’s possible with Claude to build great products for our customers.
Claude: Michele Catasta (@pirroh) is President and Head of AI @replit, the platform where anyone can build software in natural language.
At 16, he set out to make software open to everyone. Today, over 50 million people are building on Replit with Claude:
biohub
🎙️@alexrives on 'AI for Science' with @latentspacepod breaking down our world model of protein biology: ESMFold2, ESMC, and ESM Atlas. https://www.latent.space/p/esmfold2
bring-your-own MCP servers:
OpenAI Developers: Private MCP servers 🤝 OpenAI products
Your team can keep MCP servers inside your network while ChatGPT, Codex, and the Responses API connect through outbound-only HTTPS.
🔗 https://developers.openai.com/api/docs/guides/secure-mcp-tunnels
Tymofiy Mylovanov
Petraeus: Putin is the personification of evil.
He wants to reassemble as much of the former Soviet Union or Russian Empire as he can, and he does not believe Ukraine has a right to exist as an independent country. We did not listen carefully enough. 1/
Codex for parallel browser-using subagents:
George Pickett: The wild part of Codex sub-agents isn't that one AI can use Chrome.
It's watching a single prompt turn into seven browser sessions running at the same time.
Flights, cars, Airbnbs, hikes, forms, checkout pages.
Still rough around the edges but still feels like the future
Dr. Catharine Young
In case you still think there is any normalcy in how scientists are being treated at NIH.
Photo credit: 27UNIHTED
Brycent
YC is backing some absolute wizards!
Garry Kasparov
Yes, it was easier to laugh, or cry, at such lunacy in medical and scientific officialdom in banana republics like Russia before Trump appointed RFK Jr (circumcision and vaccines cause autism, mRNA tech is evil, etc.) and 52 GOP Senators voted to confirm him.
Branislav Slantchev: Russia’s descent into madness continues as all sorts of crackpots and grifters take advantage of the marginalization of science and politicization of everything, including medicine. Here, at the official meeting of the psychiatric association, they are saying people are being
Liz4SF
Zero common sense, let’s do away with SATs bc it’s racist. UC regents should all be fired, along with/ racist admissions committees, where Asian kid w/ 4.42 Weighted GPA & 1590 on SAT cant get u admissions to UC SD, UC SB, Davis nor Irvine
“More than 500 professors signed an open letter saying they’ve been forced to teach “middle school” math in Calculus and other courses as incoming students are alarmingly unprepared for college-level coursework.”
New York Post: UC professors forced to teach 'middle school math' after SAT ban https://trib.al/VJYmde8
Absolutely it is
Rexan Wong: NOW IS THE BEST TIME IN HISTORY TO START A COMPANY
heres what the CEOs of @stripe + @Replit just said:
> new business creation on stripe is up 2x YoY in march (even bigger spike than covid lockdowns)
> 20% of startups charge their first customer within 30 days (was 8% in 2020)
Kenneth Roth
Trump is now "the most unpopular president since our poll started in 2009. His net approval rating is -24," due to his war of choice in Iran and his mishandling of the economy.
https://trib.al/G1c4KcW
Occupy Democrats
BREAKING: Trump just got busted STEALING all the donation money from his “Board of Peace” fund to rebuild Gaza!
This one REALLY takes the cake...
When Donald Trump announced he would be creating a new multinational organization to compete with the United Nations and manage the colonization of Gaza, it was roundly disparaged as being an obvious attempt to solicit yet more bribes from other countries.
These concerns turned out to have been well-founded, as a new Financial Times report shows that the Board of Peace fund is empty.
“Zero dollars have been deposited,” one source said. “Member states” pleged $7 billion while Trump himself promised to put $10 billion in.
At least…not to the official World Bank fund.
Instead, donations have instead reportedly flowed through a separate JPMorgan account controlled by the organization. Unlike the World Bank fund, the account is not subject to the same transparency requirements.
Meaning he’s stealing all the money, which everyone figured he was going to do in the first place.
While nobody believed that Trump had any intention of actually rebuilding Gaza, it is still shocking to see him perpetrate such a wide-scale and brazen theft in front of the whole world.
Every single cent needs to be stripped from him and given back to the people of Gaza once Democrats are back in charge.
Ankit Gupta
my prompt for aspiring founders: assume the models reach superintelligence (arguably they have already), don't kill us all, and still require us to prompt them to do things.
what are the hardest problems you can now point them at?
Underappreciated how capable GPT-5.5 is at cybersecurity:
Philo Groves: GPT 5.5 found a 27-year-old RCE introduced in April of 1999. I've triple-checked the flow and commit history, it's real.
Can't wait to responsibly disclose!
Bring back the SAT at the UC system
(yes that's correct, they don't even look at SATs anymore for college admissions there)
Neetu Arnold: University of California STEM professors want standardized tests back due to severe math deficiencies among students:
“We now observe preparation gaps so severe that instructors must reteach middle school mathematics”
“The current admissions metric, based primarily on GPA &
It's time to make new software, not recreate old software
When the form changes, what you create must change
Aaron Epstein: A lot of people are worried software is dead.
But throughout history, any time humans have gotten new tools, we've always found new things to create with them that weren't possible before.
It just doesn't always look exactly like the things of the past.
Latent.Space
🆕Biohub’s Protein World Model: ESMC-6B, ESMFold2, 6.8B proteins, 1.1B structures, antibody design, SAEs, & the bitter lesson for biology https://www.latent.space/p/esmfold2
@biohub Head of Science @alexrives explains why biology may scale like language modeling, how metagenomics unlocked the next ESM scaling curve, why protein LMs can learn structure/function from sequence alone, how sparse autoencoders reveal biology inside the model, why ESMFold2 can beat specialized systems on antibody-antigen prediction, and how Biohub’s $500M Virtual Biology Initiative aims to build predictive models of cells, disease, and eventually physiology.
OpenAI for self-improving tax agents:
Samay: At @ThriveHoldings, we built a product with @OpenAI to automate tax prep for the 30+ accounting firms we own across the country.
This season, it processed 7k+ returns. But what I think is more interesting is that the product meaningfully self-improved as accountants used it.
