
We asked Claude to design three electronic devices: a speakerphone, a voice-assistant hub, and a handheld virtual pet. The prototypes worked without any fixes to their circuitry.

We asked Claude to design three electronic devices: a speakerphone, a voice-assistant hub, and a handheld virtual pet. The prototypes worked without any fixes to their circuitry.
For our last movie night we watched Princess Mononoke.
a trip report

For Claude to learn from experience, it has to decide what’s worth remembering before it knows what’s next. Here’s how “dreaming” helps it organize memory and correct mistakes.
From the archive
↗GPT‑6 is rolling out globally in ChatGPT with Intelligent UI, delivering faster responses with visuals and interactive experiences you can explore and use…

Our house has a box room, full of boxes that we still haven't unpacked from our last move.

AI Science startup converts a quantum conjecture into a theorem:…An evocative taste of the singularity…In the same way a few years ago it went from remarkable…
Agent access without exposing your network — Tailscale OAuth apps are now GA.
From the archive
↗
Our Series D round of financing.


Most of the changes I make to my projects these days come in the form of PRs opened for me by Claude Code or Codex.

Welcome to the October edition of the NeurIPS monthly Newsletter!

ARI: a retriever that updates its own weights during search, using gradients derived at inference time from the judgments of a reasoning model.
In middle school I could not get through long passages no matter how hard I tried.

AI for Science grantees Jakob Macke, Daniel Gedon, and Stefan Wahl on how AI is starting to change how scientific models are built and interpreted.

Nvidia CEO Jensen Huang admitted to an affair with a Chinese spy, and that might not be the end of his troubles.

We find that frontier LLMs are surprisingly good at natively reading and writing neural network weights, using a variety of recent tools and algorithms or even via direct weight editing.

October 16, 2026 began like any other day at Fly.io (meaning, we were anticipating only a medium amount of chaos).

AIs are now good enough to run a conspiracy

Today, we’re inviting members of the life sciences community to apply to work with us on special projects.

Learn how OpenAI’s Control Room tracks model workloads across our research infrastructure and can rapidly stop them when safety or security concerns arise.

How we made clusters scale up on demand and resize without downtime, and rebuilt our control plane along the way.


In this report, we use privacy-preserving infrastructure to study the scientific use of Claude, Claude Code, and Claude Science. We document how usage varies across scientific fields and research tasks, identify what explains these usage patterns, and outline scenarios for how AI could accelerate research.
ChatGPT Enterprise is now available through AWS Marketplace, helping organizations simplify procurement through existing AWS purchasing and billing…

Diffs, Linear’s fast and focused code review, is now on iOS and Android.

Modern post-training brute-forces ever more environments and ever longer horizons, because frontier Transformers are still poor at compositional generalization. A good harness makes every model call locally in-distribution, so what it learns carries across lengths and domains.

govern and trust AI-generated code and Codex to ship 20 times more software with the same engineering team.


We’re sharing another finding from Anthropic’s biology research lab: the first 3D structure of a vault particle found in prokaryotic cells.
Yesterday's hot paper was Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights by Rikiya Takehi, Jayden Teoh, Alex Zhang…
Kuken Industries uses the OpenAI API to search 70 years of cooling tower drawings.
My first blog post for Antithesis is now up!

Instead of YouTube, watch HackerTube!



For contest #527, we explore a city dominated by rivers and early explorers.
Our first open model economy research note for paid subscribers.
AI could create a billion new programmers.

16 papers accepted to the Machine Learning Reproducibility Challenge (MLRC) 2026, an official NeurIPS 2026 track.

It's truly the best option for you, your party, and the country.
How Neki detects a failed Postgres primary, and chooses a new one from the WAL that survived.
When we looked at how well agents used test/verification techniques in August 1, we saw that GPT-5.6 Sol generally did what you might call a caricature of…


He shares an encyclopedic knowledge of US elections.


I was reminded today of AI Adoption Rate Trending Down for Large Companies, a chart produced by Apollo Global Management in September 2025, which appeared to…
Modal powers inference for the open-weight models that Snorkel is evaluating as well as GPU and CPU sandboxes for millions of simulations per month.


Antithesis now streams test results, so you or your agents can debug within seconds of finding a failure.

November 2, 2026, including AI on the ballot, a Google state of the union, and China.
Over the last few years, LLMs have changed how software gets built and that’s no different for us at Jane Street.

Introducing GPT‑6 Cyber, OpenAI’s cybersecurity model for finding exploitable vulnerabilities and developing fixes, trained to stay within authorized scope.
two ways to consume it: YouTube, and here on Stratechery

LLMs are very bad at low-latency tasks:…Realtime-Bench says it’ll be a while till AI systems can play games or do online auctions for you…Researchers have…


OpenAI will acquire Pierre Computer Company, expanding Codex’s tools for building, understanding, and reviewing code.


Network-attached storage is a fine choice and a bad default.

OpenAI is partnering with Hacktron to find vulnerabilities, test connections with third-party services, and strengthen defenses with frontier AI.

An interview with Intuit CEO Sasan Goodarzi about the working for one of tech's oldest software companies, and how Intuit is leaning into the AI paradigm shift.


A small-scale rule of thumb.
OpenAI introduces BioBasicsBench, a 100-task benchmark measuring how well AI agents interpret raw experimental data and handle everyday wet-lab science.

Data Centers were on the ballot, and they lost in a major way; what does it mean if American opposition to building is permanent?
"A lot of the stuff that I cut out, it's tragic that no one knows it."
The apartment is packed up and empty and I'm leaving New York for Durham this Saturday.
OpenAI built an autonomous pipeline that found, validated, and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel.