2026-06-20

"if UC chooses to resist (or appears to resist) the pressures to change the current admissions procedures, outside political forces might compel change, and mandated reforms likely would be less informed, less effective, and more prone to produce unexpected and undesired outcomes than reforms produced within the University" -- UC Standardized Testing Task Force Report, Jan 2020
Who were these mysterious "outside political forces"? And how could they "compel change" when UC's autonomy is written into the CA State Constitution?🤔

I used to be a die-hard Claude Code user.
Codex has won me over because:
→ GPT-5.5 is excellent
→ Fast mode + generous limits = more reps
→ Little touches like steering, auto remote control on phone, etc
But most of all Codex's browser and computer use capabilities are simply goated. I built so many workflows relying on those two things alone instead of hunting for APIs.
I still use Claude Code too. The app seems to be getting better and the design and frontend capability of Opus is still much better than GPT. Whenever Fable comes back that's another reason to go back.
Honestly, I hope these two compete forever and other players (Cursor/Grok, Gemini, etc) all stay competitive.
This way the builder keeps winning 🙂

This guy collected every local law in America and put them in a single database.
2.2 million laws.
This might seem to you like some nerds side project, and while it is technically, it is also much more important than you think.
Historically, whenever anything is made transparent to the voting bloc it undergoes radical change.
Every. Single. Time.
To the extent governments work overtime to make things opaque. (And I expect that to happen to this dataset too very soon, in a thousand different ways.)
The internet makes everything including politics “global/local” and local laws won’t escape.
It’s hard to understand right now how profound this will be, but we live in interesting times for sure.
Thank you Joe, and all the other nerds who do stuff like this because they are personally curious and they can. This is how history moves forwards
Joe Barrow: New paper: every law in America is technically public. But not really, until now!
With @DenisPeskoff at UC Berkeley, we built a corpus of ~every publicly accessibly city and county law, and released a huge chunk of it!
2.2 million laws, you're (probably) covered in it!
🧵

Vibecon recap

If this surprises you, you probably didn't know Ben Gvir was at a wedding where they danced to the hit Israeli song "Ali Is On The Grill", which celebrates the burning alive of an 18-month-old Palestinian child, while stabbing his picture
איתמר בן גביר: על כל דמעה של אמא ישראלית, אלף אמהות לבנוניות צריכות לבכות. לבנון כולה צריכה לבעור!
עם כל הכבוד לאמריקאים, ישראל חייבת להבהיר לעולם כולו שדם בנינו וביטחון אזרחנו איננו הפקר. לבנון כולה צריכה לבעור. חובתנו העליונה היא להגן על אזרחי ישראל ועל חיילי צה״ל, והמחויבות הזו קודמת לכל

Looks like @Replit is coming to the UK soon 👀
A win for London!

I have read approximately 3,000 10-Ks in my life. I have read my wife’s emotional state correctly maybe 11 times. This is troubling because the skills should transfer. Both require you to look past the headline. Both require you to read the footnotes. Both require you to notice what was said last quarter that is not being said this quarter.
I can spot a goodwill impairment from 40 pages away. I cannot spot that my wife has been quietly furious since Tuesday. In a 10-K I notice when management changes the word “challenging” to “dynamic” and I correctly interpret this as a warning. In my marriage my wife changed the word “fine” to “fine.” and I did not notice the period. The period was the entire disclosure.
I missed it. I read a footnote last week in a packaging company’s annual report that disclosed a related-party transaction worth $400,000 and I caught it in 90 seconds. My wife told me three times this month that she was tired and I interpreted this as “tired” when in fact it was a Level 3 disclosure requiring immediate management response. I have a system for 10-Ks. I read the MD&A first, then the risk factors, then the cash flow statement, then the notes. I have no system for my wife. She is a company that does not file. She reports continuously and without warning and the format changes every quarter. Her risk factors are not enumerated.
Her MD&A is delivered through sighs of varying length and I have not yet developed the ear. Last week she said “do whatever you want” and I did whatever I wanted and it turns out the correct interpretation of “do whatever you want” was “do not do that specific thing” and I have no idea how I was supposed to know that, and yet, looking back, the signals were all there. The signals are always there. I have been trained to find signals. I find them in companies I will never meet. I miss them in the person I have lived with for nine years. My wife has started saying things like “you would notice this if I were a stock” and she is correct. She is correct. If she had a ticker I would have already built a 6,000-word model on her. I would know her seasonality.
I would know her capex cycle. I would know which quarters historically run hot. Instead I treat her like a private company and I am surprised every time the auditors arrive. I am going to bed now. She said good night in a tone. I do not know what the tone meant. I will find out in the morning. Or I will not. The 10-K of my marriage is filed in real time and I am, as always, three quarters behind.

You can now make videos like this using HTML.
I didn’t believe it at first until the HyperFrames team showed me how they do it directly in Codex and Claude Code.
Best of all, the tool is 100% free and open source.
But you need to learn the right workflows to make the videos look great - including generating a frame.md, storyboards, and more.
📌 Subscribe to get our full tutorial tomorrow: https://www.youtube.com/@PeterYangYT?sub_confirmation=1

I’m old enough to remember this being called “vividly overrepresented”
The Athletic: Half of the U.S. men’s national team is Black. After decades of overwhelmingly White teams, the makeup of this team is powerful, writes @HenryBushnell.

Re @garrytan was told he was crazy for saying this months ago.
He was ahead
Guillermo Rauch: The next hot programming language is… markdown.
A minimal eve agent:
📂 𝚊𝚐𝚎𝚗𝚝/
📄 𝚒𝚗𝚜𝚝𝚛𝚞𝚌𝚝𝚒𝚘𝚗𝚜.𝚖𝚍
📂 𝚜𝚔𝚒𝚕𝚕𝚜/
📄 𝚢𝚘𝚞𝚛-𝚎𝚡𝚙𝚎𝚛𝚝𝚒𝚜𝚎.𝚖𝚍
Deployable in one command: 𝚟𝚎𝚛𝚌𝚎𝚕.
It’s the most accessible programming has ever been. And

🚨🚨🚨A research project idea! How to measure world models?
Everyone's talking about world models these days. World model here, world model there. We can argue about what "world model" actually means, and we have some interesting results on that, but let's assume some hand-wavy definition for now. The real question is how you tell a good world model from a bad one.
Here's my bet on what actually matters, and it isn't how real the video looks or whether you can read off physical quantities from some giant latent. It's the structure of how the latents evolve. Say a rock is flying towards you and you want to plan an escape — do you care about its velocity? its energy? the projection of its velocity onto the 2D plane? They're all nonlinearly related, so chasing any one of them is the wrong target. What matters more is whether there's a simple, interpretable rule for how the latents move. And it has to be the latents — the real degrees of freedom aren't the pixels. Write down the map from past latents to future latents: what's its Jacobian? The world is noisy, so do errors propagate in a reasonably uniform way, or are there points where a tiny error blows up immediately? And separately, find the smallest latent space that still works and check whether its dimension matches what physics says it should be.
Meanwhile the field measures world models in a dozen ways, and they quietly disagree with each other. Representational probing puts a linear probe on the representation and asks what's decodable. Information-based methods try to measure something like the predictive information the latents carry about the future. Rollout error runs the model forward and tracks how fast it drifts from reality. Then there's downstream task success — does it help on some task — versus closed-loop utility, where an agent actually plans with it and you see whether it succeeds. Different things, and they often disagree.
So two ways to make sense of the mess. One is empirical: take many settings and many models, measure all these metrics, and map how they correlate, similar to the "Fantastic Generalization Measures and Where to Find Them," paper, but for world models. The other is to stop treating them as separate benchmarks at all. There's really one thing underneath: a system that changes over time, that you can act on, and that you only ever see part of. Every metric is just one shadow of it — realism checks the frames, rollout error checks the predictions, probing checks the state, and closed-loop checks the planning. Each hides an assumption about when it's even a fair test. Make those explicit, and ask the real question: when does doing well on one metric actually guarantee doing well on another, and when do they just happen to agree?
I'd love to hear what people think, and let me know if you want to collaborate on it.

Notes from the red hot San Francisco apartment market:
- almost no supply, across neighborhoods
- tenants regularly offer to pay above asking rent, pre-pay X months
- 90% of new tenants are moving in from out of town
- heard a story of someone inquiring about 50 apartments, received responses from less than half
- parents not just co-signing, but paying rent for their kids
- waiting lists now common, so many units are not publicly marketed
- rents seem high, but people make so much money affordability also still high (not just tech workers, finance, legal etc)
- turnover of existing units down, embedded loss-to-lease up
San Francisco is feeling the impact of decades of under-building, especially in the neighborhoods.
Market rate projects still don’t pencil due to high building costs and a complex entitlement regime thanks to multiple competing regulatory frameworks.
Is it state density bonus? Builders remedy? Family zoning plan? What will change once I submit plans?
Why would capital build when they can buy existing at better yields and half the basis.
These conditions could persist, for longer than most think possible.

As India builds frontier AI rooted in its own languages and knowledge, IIT Bombay and BharatGen are proud to support and participate in Project Tapestry: an open, global consortium for nations to advance frontier AI together. BharatGen, supported by the IndiaAI Mission and the Department of Science and Technology, joins as a founding contributor with multilingual AI infrastructure built natively for India's diverse languages.
A Letter of Intent was signed on June 18th by IIT Bombay Director Prof. Shireesh Kedare, in the presence of both Deputy Directors Prof. Milind Atrey and Prof. Ravindra Gudi, and Prof. Ganesh Ramakrishnan, Founding Director at BharatGen. Project Tapestry was represented by Dr. Christopher Nguyễn, Chief Architect of Project Tapestry and AI Alliance Board Member. What India has built for its own people, it now brings to the world.
@EduMinOfIndia
@BharatGen_Com
@ganramkr
@OfficialINDIAai
@IndiaDST
@ylecun
@kb_bha
@pentagoniac
@kb_bhatta

In 1958 Ian Donald published what is now the foundational paper on medical ultrasound for obstetrics. He was so widely ridiculed by his colleagues at the time that they nicknamed him Mad Donald [1], and one said ultrasound would be useful only to "a gynecologist who was blind and had lost the use of both hands" [2]. Another noted that he'd invented a £10,000 device to undertake a task that could be accomplished with a £0.02 rubber glove. [3]
Last week, my wife and I welcomed our first child into the world. She had a rare pregnancy complication that until recently would have meant only a 28% intact survival rate for our newborn. But in 2013 US guidance was updated to add regular preventative screening for her condition at the 20-week ultrasound, and with early detection the survival rate is ~99%. (In the UK, preventative screening is still not recommended, for reasons like "it is not known how accurate screening tests are" [4].)
The entire history of radiology is people expressing skepticism about the work done by innovators. I for one am grateful for folks like @DavidSHolz building new classes of devices that can help us see things in new ways, and I'll be rooting for their success. Hand in hand with my wife and our healthy baby boy.

For those who feel like Beginner AI Builders 👷🏼:
I want to share something I’m launching next week. This could be for you or someone you know who’s just felt overwhelmed by the whole AI thing and feels like it’s passing them by.
Each day next week is a short step in getting a To Do list on the web up-and-running and fully functional.
It’s easier than people think and I want to show you how.
The article at the top of page below will explain more and, if you subscribe, you’ll get a short email with step-by-step instructions each day next week.
Starts Monday so subscribe now so you don't miss out.
http://alextalksai.kit.com

when Vincent superimposed last year's floor plan ontop of this year's we were all quiet for a little bit
lily: just noticed AIE 2026 is at SF Moscone Center this year, huge upgrade from Marriott Marquis. we still love the old venue, but Moscone hits differently. @swyx @aiDotEngineer.

BREAKING: Donald Trump is reportedly furious that the annual ranking of US presidents by a committee of 50 top presidential historians was released today, and the committee almost unanimously ranked Trump’s second term as the worst in US history.

1.3 GW secured. 6 GW needed. The gap is the story.
Anjney Midha @AnjneyMidha runs AMP, a grid operator that pools AI compute for frontier labs.
His own pipeline shows 1.3 gigawatts secured against an estimated 6 GW needed over four years. The 4.7 GW gap is years of constrained supply for independent labs. Anticipated 2026 year-end excess vanished in six weeks.
If you model AI infrastructure demand, this says the shortage is structural, not a 2024 hangover.
How a compute grid actually clears supply and demand, on Latent Space @latentspacepod: https://podcastalpha.substack.com/p/why-ai-labs-with-unlimited-gpus-still-fail
Source: Latent Space - https://www.youtube.com/watch?v=h5dlIPM0X18

Re it looks like AIE 2026 is mostly sold out, for anyone who couldn't get a ticket for this year, I love that AIE would livestream and have this archive of the conference and workshop.
A few of my favorite talks from AIE 2025.
1. @bcherny, Anthropic, Claude Code
2. @JustinLin610, Alibaba Qwen: A Generalist Agent Model
3. @hwchase17, LangChain/LangGraph
4. @danielhanchen, UnSloth, Reinforcement Learning

They’ve arrested more people for touching the peeling paint at the Reflecting Pool than they’ve arrested from the Epstein files.

Anthony Fauci is an American hero who kept us protected from deadly viruses for decades. The people spreading lies about him are spreading Russian propaganda because they get paid to do so. They are the real traitors. Russia would love to know what we have in our biolabs so they can use that research to hurt Ukraine who is dominating them with drones.

Re Update! Thanks to @davidhoang and @swyx I have purchased my ticket 🙏 see you all there!

Interactive AI movie site built in Replit with three short films using AI video generation and vibe coding. Powered by Flora AI and Kling AI.
Gryun Kim: 🎬AI Movie Website🎬
👉 http://gryun-aimovie.com
interactive website where you can explore 3 AI-gen short films made entirely with 'AI video + vibe coding'
experience a message about direction in the AI era
Powered by @Replit @floraai @Kling_ai

This Japanese restaurant logo is simultaneously a Japanese crane and a hand holding chopsticks.