Keyboard shortcuts
The people building AI, writing on their own blogs. Last 30 days.

a quiet day lets us pilot a new show
New research shows that squid lived all around the globe even before the Cretaceous-Paleogene mass extinction event. As usual, you can also use this squid post to talk about the security stories in the news that I haven’t covered. Blog moderation policy.

The Firefox team are asking us to weigh in on proposals for Interop 2027—what features should browsers be concentrating on implementing or improving. Here are my top five: Web app installation via <install> and navigator.install(). Prompts the user to install a progressive…

The best Stratechery content from the week of October 12, 2026, including what Apple is doing, SpaceX's appetite, and the title, takes, and the Washington Wizards.

Hot on the heels of reading and enjoying The Mountain In The Sea, I picked up Ray Nayler’s next work, this tightly constructed novella. Once again it deals with inter-species communication. This time it’s de-extincted mammoths. Once again, it’s set in the relatively near future.…

Getting the results back out. The post C++ futures part 5: <CODE>std::<WBR>future</CODE> and <CODE>std::<WBR>shared_future</CODE> appeared first on The Old New Thing.

AI might not yet be eating the physical world. AI is still eating the world of news, and I am finding less and less opportunity to post about other things. My backlog of items grows across most of my secondary … Continue reading →
This is a problem: Any open model whose weights have been edited can carry a backdoor, whether it’s a task-specific fine-tune, a merged adapter or an abliterated build. We have abliterated models in the title because it’s the most popular reason people download modifi…

Nick Cammarata out of nowhere with some major claims!

If AI is so great, why aren’t we seeing an explosion of new, amazing software products? If anyone can turn their idea into working software with coding agents, where is all that software? People have been asking this question for years. As far as I can tell, the answer is that it…

The amount of code being written at Anthropic continues to rise exponentially. It is being managed by Scion, their internal replacement for GitHub that is better suited to the scale and style of the operation, which they created with a … Continue reading →

Back in 2019, I read the book Other Minds: The Octopus, the Sea, and the Deep Origins of Consciousness by Peter Godfrey-Smith and said: I really enjoyed this evolutionary tale. It’s equal parts biology and philosophy. I will never look at cephalopods quite the same way again. Rea…

ASML is building a hybrid bonding tool! The company has not announced anything yet, but its patent filings, a news report and its own roadmap all suggest so.

curl started out GPL licensed, as direct inheritance from its httpget days, but in October 1998 we switched to the Mozilla Public License (MPL). In 2001, we added the MIT license as an additional license to optionally use in addition to MPL. We made curl dual licensed because cle…

Also: insider details on how and why Anthropic moved off Git, a new trend of AI-assisted coding rounds inside of Big Tech, and more
Matthew Green has a new site that tries to estimate the number of cryptanalyst-hours that have been spent analyzing different public-key algorithms: Abstract: We estimate how much public cryptanalysis each public-key assumption has received, using the literature itself as the ins…
Providing the case that the C++ language won't do. The post C++ futures part 4: What if I want an immediate on-thread <CODE>std::async</CODE>? appeared first on The Old New Thing.

When software meets hardware, engineering principles remain largely the same, but the sensible defaults often change. Kuldeep Singh describes our experience modernizing a large IoT fleet of battery-powered sensors. Our familiar principles remained important; what changed was how…

Sandboxes are all you need.

A conversation with Hollywood super-agent Ari Emanuel about his memoir, the impact of AI on Hollywood, and the power of a phone call.

JPEG XL has landed, but how does it compare?

How do we know the company’s stated reasons for firing three safety researchers are its actual reasons?
Uber Distinguished Engineer Uday Kiran Medisetty shares how Uber builds internal developer tools, integrates AI agents into engineering workflows, and measures their impact on productivity.

We need the frontier labs to have a deep and robust safety culture. Alas, this is not the case. Anthropic devotes considerable effort to cultivating and maintaining such a culture. It shows, especially among the members of technical staff. Having … Continue reading →

Building a slide template starts with three foundational choices — colours, fonts and layouts. Sumeet Gayathri Moghe covers the basics of color theory and how to build a reusable palette, the essentials of typography, including serif, sans-serif, handwritten and monospace…
Doing things later in different ways. The post C++ futures part 3: The mental model for <CODE>std::async</CODE> appeared first on The Old New Thing.

The Apple Homeview is official, and the question is how big is the addressable market for expensive devices based on subpar AI.

Over the last year or so, multiple engineers at Aisle have spent a lot of time and energy while pointing their AI powered tooling at the curl source code. This has resulted in no less than thirty-five published curl vulnerabilities and CVE Ids. Today, we make Aisle’s audit…

In the curl 8.23.0 release we shipped a fix for one of the worst vulnerabilities in curl in a long time. We grade it severity HIGH. It is known as CVE-2026-92392 with the title HTTP/2 upload Use-After-Free. This post explains the bug, what the problem can lead to, and something a…

I am a habitual tab hoarder so I created a Firefox addon to help me manage every one of my very special browser tabs. Sort tabs by oldest, unload them, find duplicates, and consolidate tabs to a single window — buncha useful stuff in there! Download from Mozilla Addons Source cod…

In the hours following the release of CVE-2026-92392 for the project curl, site reliability workers and systems administrators scrambled to desperately rebuild and patch all their systems to fix a use-after-free in HTTP/2 upload retries where cleanup turns leftover upload data in…

How to handle uncertainty with AI vendors and disaster recovery (DR) approaches to consider. An excerpt from the book Building Resilient Distributed Systems by Sam Newman
A couple of months ago Jim Highsmith explained how practitioners writing for each other forms a distinct style of writing. He now adds a section to the article with more detail on the craft of how we write in this style. How the writer shows up in the prose, its four principles,…
I’ve started working on a game about being a frustrated and ineffectual manager of a goblin village. Most of the work is writing narrative snippet and then asking Claude to turn it into code. On the phone, usually on the way back from school drop off.

Python 3.15 is here! Here's the essential What's new in Python 3.15 guide, this year curated by 3.14 and 3.15 release manager Hugo van Kemenade. As someone who builds open source software that works in every supported Python version, I get to say goodbye to Python 3.10 and start…

Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now New benchmark shows how AI models lack intuitive visual reasoning:…Humanity’s Sixth Sense highligh…
Many software engineers are righteously angry about AI. It’s understandable to be upset that the skills you built your professional identity around are getting automated away. But software engineers are the last people in the world who should be complaining about this. If you liv…

Atlas Infinite, which dropped in public preview on September 29 alongside MongoDB 9.0, delivers a significant architectural update payload: compute-storage disaggregation. Separating compute from storage is not new, and there are well-known ways to do it. But what makes Atlas Inf…
When you disagree with someone, there are two paths forward. The first is to try to engage with what they’re saying: what their reasons are, whether their position is true, what evidence could be brought to bear, and so on. The second is to diagnose them: to engage with the secon…
TL;DR? Jump to demo I’ve always had a soft spot for icons 1. Not only do they add polish to a UI, they can also tangibly improve usability. Familiar icons help users find what they need quickly, since we recognize pictures much faster than we read words. And as a web architect at…

I’ve often been surprised when I hear from top researchers in industry that they think AI will be better than them at their job in a few years, and I didn’t really know why I doubted it.
Have you tried to explain lobbying to a non American unfamiliar with the concept?

I shipped a new feature for my blog today: the Newsletters page, which offers an index of all of the newsletters I've sent out, both my free weekly Substack and my monthly sponsors-only updates. I built the feature almost entirely using my voice, chatting away to my laptop while…

Brilliant jerks, ZIRP-era managers, and how psychological safety lost the plot. Part 3 of my conversation with Dr. Cat Hicks.

AI safety is on fire.

We have a choice in front of us. In his latest op-ed for Transformer, Yoshua Bengio invites employees of frontier AI companies to ask themselves if their work is truly leading us toward a safer future. Reflecting on these questions is hard, but we no longer have the luxury of tim…

OpenAI, CloudFlare, and Amazon have all copied this tiny startup.

Monte Carlo simulations (or methods ) is the technique of applying randomness and the Law of large numbers to the solution of various scientific and engineering problems. One of its first documented uses was by Stanislaw Ulam and John von Neumann for nuclear weapon simulations af…

As previously promised, here's Anthropic's new fast, low cost model: Introducing Claude Haiku 5.5. The previous Haiku, 4.5, was very much showing its age. It came out almost a year ago, and was priced at $1/million input and $5/million output - relatively expensive even back then…

In the 18th century, 90% of Americans worked in agriculture to feed themselves and the nation. Today, it's around 1%. In the 1950s, there were hundreds of thousands of telephone switchboard operators. Today, they're all gone. There used to be well over a million professional typi…
I have recently improved my mental model of Git. Consider these two git commands:

Open-weights, ideology, and acknowledging trade-offs.
Two weeks ago at Rails World, I told my fellow programmers that it's time to put down the pencils. We're not going to write the vast majority of code by hand any longer. Coding agents have gotten so good that it's simply not an economically viable enterprise going forward to have…

Visiting San Francisco for the first time? Welcome. The thing to understand about SF is that it is a city of duality. It is probably the only city where you can see a self-driving car drive itself past a homeless person. Within a few dozen blocks, you can traverse the entire spec…

More than a year ago I wrote a few posts here that recommended people not to load custom tools into their context (or MCP servers) but to just use more scripts. Most importantly I wrote that Code Is All You Need and I wrote about that MCP needs code. With Pi 1.0 we now added MCP…

Limit testing every AI subscription plan from Anthropic, OpenAI, Meta, SpaceXSI, MiniMax, Moonshot, Z.ai, Cursor, and Cognition

Understanding AI has a new writer!

A dangerous myth has gained traction after the recent hacks perpetrated by the AI agents developed by leading companies: That these incidents are just cyber security problems, that it's only a matter of fixing the sandboxes in which they’re trained. In his latest op-ed for the Fi…
Paradise is not lost, the world is not ending, and most material metrics of humanity have never been better. Yet the endless, incessant dooming about the climate, the inequality, and AI is everywhere. We may have killed God, but we clearly didn't bury the devil's anxieties with h…
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now When should you use swarms? When you are in a hurry:…How does swarm scaling work?…Toby Ord h…
How long should a micro benchmark run? My rule of thumb is to tweak the input size until the benchmark takes about 300ms, for the following reasons:
Interesting & joyful things from the previous week
Strands Decider: Why Not an Encoder? Expertise level: exhausted. MathJax = { tex: {inlineMath: [['\\(', '\\)'], ['$', '$']]} }; As I admitted in my last post, I am very much not an expert model developer. What follows is likely to be, at least partially, inaccurate. I have tried…

FFXIV raid mechanics explained in plain language.
Here's a product feature which the world is going to need a whole lot more of over the coming months and years: default hard budget caps. I'm talking about the feature of pay-by-usage services and APIs that lets you say "after $X/month, cut this thing off and return errors". Thes…
So when I saw Dyson had a $500 toothbrush, I was excited. Finally, advertising that targets me! I love brushing my teeth, and I have more money than I know how to spend. Not because I’m particularly rich, but because most stuff doesn’t really appeal to me.

Fall of the Incas & Aztecs.

Benefitting from the Bitter Lesson
"Wuthering Heights" by Emily Brontë - good writing, but the protagonists are quite something. There's barely a likable character in the whole story - they are all either crazy, evil, stupid or a combination of these. It's also remarkable to consider how small the world…
They will come for your thoughts, and I’m not sure what your defense will be. For surely, they are better at thinking than you. That choice you are about to make, well you see, we have modeled your life, and have calculated you will be happier making this choice instead. And they…

Anthropic released new eval tooling for Claude Code. Their claude-api plugin now includes a new build_eval and hill-climb command that helps you build evals, check the graders, and improve your application against them. I usually don’t review eval tools because software changes s…

I'm at OpenAI DevDay today, in Fort Mason, San Francisco. Same as last year I'll be live blogging the keynote and some other notes during the day. OpenAI gave me a free ticket and a seat in the "creator" area for the keynote. You are only seeing the long-form articles from my blo…

A Visual Guide to RNNs, CNNs, Transformers, and Calibration, with Hands-On Experiments on Accuracy and Efficiency

Serde is an amazing serialization library for Rust and it has been a huge reason why I felt productive with it for years. However already while at Sentry I got quite frustrated with some of the limitations with it but actually replacing Serde is tricky because of the might that i…

Headlines and History The headlines keep coming: such and such an AI system has solved such and such a math problem. And more and more I’m hearing people saying: maybe we don’t need people doing math research anymore; maybe we should just delegate it all to more and more powerful…
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. Subscribe now Are minds patterns from a Platonic space, with bodies and machines as their interfaces, Michael Le…
Small Decisions: Engineering a Leading Model Or trying to, at least. Update: An evolved version of the model described here is now available as part of Strands. Check out strands-decider on github and HuggingFace. My day job has been primarily in AI for three years now, but I’d b…

Why I'd invest in post-training existing open-weight LLMs, with Fireworks' Ember-1 as an example of more token-efficient reasoning.

Hello! Recently I needed bike lights for my bike. And I remembered that I already had rechargeable bike lights that I bought ten years ago, that I hadn’t tried in a long time. I tried to recharge them, but after fully charging them, they only worked for maybe 5 minutes befo…
Like many programmers, I find Alexis King's Parse, don't validate article fascinating, because it gives a name to an idiom that seems familiar and important - one I've observed and used in the past without naming it explicitly. This post is a review of the "Parse, don't vali…
Interesting & joyful things from the previous week

Agents are great at producing code. The valuable work left for you is deciding what should exist, how it should behave, and what you're willing to trade for it.

General methods that scale with computation will inevitably displace hand-engineered domain knowledge! Sutton’s Bitter Lesson hangs like a Sword of Damocles over all of us. This paper, which just dropped, applies that lesson to data agents, and says that as LLMs improve, they wil…

In this speech, Professor Yoshua Bengio addresses the UN Security Council to discuss the unprecedented threat posed by uncontrolled frontier AI. Humanity has summoned the courage and wisdom to manage catastrophic risks before. This is one of those moments, far bigger than any per…
Hacker News (aka HN) is a link aggregator where “hackers” 1 gather to discuss technology, politics, and anything which “good hackers would find interesting”. HN is a means by which its host, prominent startup incubator Y Combinator, projects soft power, promotes startups it funds…
Notes on MiMo-V2.6 Pro's GQA and sliding-window attention, agent training tasks, reward signals, and large RL batches.

Podcast #19

A dialogue with Dr Cat Hicks on ethics, learning, and what we owe each other. Part one: start by listening.
My quest for a principled solution for metastable failures has taken me back to my roots on self-stabilization, as this recent paper related the problem to composition of self-stabilizing systems. But, my literature search for recent work on composing self-stabilizing syste…
c’mon, try doing a breakthrough
In one of my classes I asked the question I was afraid to ask but I just needed the answer to: “Who is afraid of not getting a job after graduating?” About eighty percent of the 150 people in the room raised their hands. That is roughly 120 students answering, in one…

.iframe--card { margin-block: 15vh; } Millions of Large-Language-Model -generated words have been spilled on craftsmanship in this gilded-age of software development. We now have agents to plan how code will be written, agents to write code, agents to review code, agents to orche…

A meditation on gaming, community, and the purpose of negative space
The other day I was rereading my Thermopylae poem and got to thinking seriously about the information dynamics of history and how these might evolve in the limit. It turns out we can go quite far in this thinking from some very basic premises and it relates somewhat deeply with..…
Interesting & joyful things from the previous week
An excellent feature of Eleventy 1 is the official PrismJS-based syntax highlighting plugin, which adds painless build-time syntax highlighting. It keeps 11ty sites from worrying about client-side bloat from JS -based highlighters and the common failure mode of over-including lan…
Are generative (randomized) tests significantly more effective than example-based unit-tests at discovering bugs? There's an interesting discussion about this on lobste.rs. One argument in favor of unit tests is, paraphrasing

Using your deep knowledge, wide knowledge, taste, and agency

This document curates the most common questions Shreya and I received while teaching 5,000+ engineers and PMs AI Evals. Warning: These are sharp opinions about what works in most cases. They are not universal truths. Use your judgment. How to use this FAQ Browse the questions tha…

You can just prove things, apparently.
In early 2025, I started seeing people turn off their brain as they use LLMs 1. They would have an LLM take an action (summarize text, write some code, etc.), and just assume that it worked 2. This generally didn't work in early 2025 and the result was often quite silly. As LLMs…

“We never want to be in a situation again where we underestimate the AI.”