What we're reading.

[latent space]

[AINews] Thinking Machines' Native Interaction Models - TML-Interaction-Small 276B-A12B - advances SOTA Realtime Voice and kills standard VAD

well done, Team Thinky.

[hacker news]

Claude Platform on AWS

[simon willison]

Thoughts on GitLab's workforce reduction" and "structural and strategic decisions"

GitLab Act 2 There's a lot going on in this announcement from GitLab about the "workforce reduction" and "structural and strategic decisions" they are making…

[hugging face]

Building Blocks for Foundation Model Training and Inference on AWS

[hacker news]

I let AI build a tool to help me figure out what was waking me up at night

[hacker news]

If AI writes your code, why use Python?

[simon willison]

Quoting James Shore

Your AI coding agent, the one you use to write code, needs to reduce your maintenance costs. Not by a little bit, either. You write code twice as quick now?…

[simon willison]

Your AI Use Is Breaking My Brain

Your AI Use Is Breaking My Brain Excellent, angry piece by Jason Koebler on how AI writing online is becoming impossible to avoid, filtering it is mentally…

[simon willison]

Using LLM in the shebang line of a script

TIL: Using LLM in the shebang line of a script Kim_Bruning on Hacker News : But seriously, you can put a shebang on an english text file now (if you're…

[hacker news]

Bild AI (YC W25) Is Hiring Founding Product Engineers

[simon willison]

Learning on the Shop floor

Learning on the Shop floor Tobias Lütke describes Shopify's internal coding agent tool, River, which operates entirely in public on their Slack: River does not…

[openai]

How ChatGPT adoption broadened in early 2026

ChatGPT adoption surged in Q1 2026, with fastest growth among users over 35 and more balanced gender usage, signaling broader mainstream AI adoption.

[hacker news]

Google says criminal hackers used AI to find a major software flaw

Unlocked:

[openai]

OpenAI Campus Network: Student club interest form

Join the OpenAI Campus Network—connect student clubs worldwide, access AI tools, host events, and build an AI-powered campus community.

[openai]

How enterprises are scaling AI

How enterprises scale AI: from early experiments to compounding impact through trust, governance, workflow design, and quality at scale.

[vercel]

Automate progressive rollouts with Vercel Flags

You can now use to roll out a feature to a growing percentage of users on a schedule, with progressive rollouts.Vercel Flags Unlike weighted splits, which hold…

[openai]

OpenAI launches DeployCo to help businesses build around intelligence

OpenAI launches DeployCo, a new enterprise deployment company built to help organizations bring frontier AI into production and turn it into measurable…

[vercel]

Vercel Sandbox firewall now supports request proxying and filtering

The now supports forwarding specific HTTP requests to a proxy you control. You can also use matchers to filter forwarding and credentials brokering to only the…

[vercel]

Node.js 26.x now available on Vercel Sandboxes

Vercel Sandbox now supports Node.js version 26. To run a Sandbox with Node.js 26, upgrade to or later, or to or later if you're using v2 and set the property…

[simon willison]

Quoting New York Times Editors’ Note

This article was updated after The Times learned that a remark attributed to Pierre Poilievre, the Conservative leader, was in fact an A.I.-generated summary…

[simon willison]

Quoting Andrew Quinn

One could say in the first quarter-century of my life, that while I was always fascinated by programming, I could never overcome the guilt of not really…

[vercel]

How Superset built the IDE for AI agents on Vercel

Superset on Vercel Software development with AI started as a single engineer chatting with a single agent about a local repo. Today, developers direct fleets…

[latent space]

[AINews] Anthropic growing 10x/year while everyone else is laying off >10% of their workforce

A quiet day lets us reflect on an interesting dichotomy in the economy.

[simon willison]

Quoting Luke Curley

WebRTC is designed to degrade and drop my prompt during poor network conditions. wtf my dude WebRTC aggressively drops audio packets to keep latency low. If…

[simon willison]

Using Claude Code: The Unreasonable Effectiveness of HTML

Using Claude Code: The Unreasonable Effectiveness of HTML Thought-provoking piece by Thariq Shihipar (on the Claude Code team at Anthropic) advocating for HTML…

[apple developer]

Brazilian betting license requirement for App Store availability

Following changes to Brazil’s fixed-odds betting regulation, apps with fixed-odds betting (gambling) features can now be distributed on the App Store in Brazil…

[github blog]

Why age assurance laws matter for developers

Youth safety requirements are moving down the tech stack to operating systems and app stores—raising new questions for open source developers. The post Why age…

[hugging face]

EMO: Pretraining mixture of experts for emergent modularity

[android devs]

Gratitude saw 25% higher retention for widget users

Posted by Ash Nohe and Amrit Sanjeev, Android Developer Relations Engineers Practicing gratitude may decrease symptoms of depression and anxiety, and improve…

[openai]

Running Codex safely at OpenAI

How OpenAI runs Codex securely with sandboxing, approvals, network policies, and agent-native telemetry to support safe and compliant coding agent adoption.

[latent space]

[AINews] GPT-Realtime-2, -Translate, and -Whisper: new SOTA realtime voice APIs

OpenAI continues deploying GPT-5 everywhere

[vercel]

Chat SDK adds Messenger adapter support

Chat SDK now supports as a chat adapter.Messenger Build agents that support messages, reactions, multimedia downloads, postback buttons, and direct…

[vercel]

Chat SDK adds web adapter support

You can now build chat UIs that connect to Chat SDK with the new . Build in-product assistants, support agents, or any other browser-based chat experience.web…

[github blog]

Improving token efficiency in GitHub Agentic Workflows

Agentic workflows that run on every pull request can quietly accumulate large API bills. Here's how we instrumented our own production workflows, found the…

[simon willison]

llm-gemini 0.31

Release: llm-gemini 0.31 gemini-3.1-flash-lite is no longer a preview . Here's my write-up of the Gemini 3.1 Flash-Lite Preview model back in March. I don't…

[github blog]

Agent pull requests are everywhere. Here’s how to review them.

A practical guide to reviewing agent-generated pull requests: what to look for, where issues hide, and how to catch technical debt before it ships. The post…

[simon willison]

Big Words

Tool: Big Words I'm using my vibe coded macOS presentations tool to put together a talk, and I wanted to add a slide with some text on it. The tool only…

[simon willison]

Behind the Scenes Hardening Firefox with Claude Mythos Preview

Behind the Scenes Hardening Firefox with Claude Mythos Preview Fascinating, in-depth details on how Mozilla used their access to the Claude Mythos preview to…

[simon willison]

Notes on the xAI/Anthropic data center deal

There weren't a lot of big new announcements from Anthropic at yesterday's Code w/ Claude event, but the biggest by far was the deal they've struck with…

[android devs]

A look ahead: Making it easier and faster to publish safer apps

Posted by Vijaya Kaza, VP, Product, App & Ecosystem Trust The mobile ecosystem is always evolving, bringing both new opportunities and new threats. Through…

[openai]

Scaling Trusted Access for Cyber with GPT-5.5 and GPT-5.5-Cyber

OpenAI expands Trusted Access for Cyber with GPT-5.5 and GPT-5.5-Cyber, helping verified defenders accelerate vulnerability research and protect critical…

[vercel]

Next.js May 2026 security release

Summary We have shipped a coordinated security release for Next.js addressing 13 advisories across denial of service, middleware and proxy bypass, server-side…

[openai]

Parloa builds service agents customers want to talk to

Parloa leverages OpenAI models to power scalable, voice-driven AI customer service agents, enabling enterprises to design, simulate, and deploy reliable,…

[openai]

Advancing voice intelligence with new models in the API

Explore new realtime voice models in the OpenAI API that can reason, translate, and transcribe speech, enabling more natural and intelligent voice experiences.

[vercel]

Vercel Flags now supports JSON values

You can now store JSON values in , extending the existing support for boolean, string, and number values. This allows you to collapse what used to take several…

[simon willison]

GitHub Repo Stats

Tool: GitHub Repo Stats One of the things I always look for when evaluating a new GitHub repository is the number of commits it has... but that number isn't…

[latent space]

[AINews] Anthropic-SpaceXai's 300MW/$5B/yr deal for Colossus I, ARR growth is 8000% annualized

And the kingmaker picks a side.

[openai]

Introducing Trusted Contact in ChatGPT

Introducing Trusted Contact in ChatGPT, an optional safety feature that notifies someone you trust if serious self-harm concerns are detected.

[openai]

Testing ads in ChatGPT

OpenAI begins testing ads in ChatGPT to support free access, with clear labeling, answer independence, strong privacy protections, and user control.

[openai]

Simplex rethinks software development with Codex

Simplex boosts software development with ChatGPT Enterprise and Codex, reducing design, build, and testing time while scaling AI-driven workflows.

[github blog]

Validating agentic behavior when “correct” isn’t deterministic

How to build the “Trust Layer” for Github Copilot Coding Agents without brittle scripts or black-box judgements by using dominatory analysis. The post…

[hugging face]

vLLM V0 to V1: Correctness Before Corrections in RL

[simon willison]

Live blog: Code w/ Claude 2026

I'm at Anthropic's Code w/ Claude event today. Here's my live blog of the morning keynote sessions. Tags: ai , generative-ai , llms , anthropic , claude ,…

[simon willison]

Vibe coding and agentic engineering are getting closer than I'd like

I recently talked with Joseph Ruscio about AI coding tools for Heavybit's High Leverage podcast: Ep. #9, The AI Coding Paradigm Shift with Simon Willison .…

[openai]

How ChatGPT learns about the world while protecting privacy

Learn how ChatGPT safeguards your privacy, reduces personal data in training, and gives you control over whether your conversations improve AI models.

[latent space]

[AINews] Silicon Valley gets Serious about Services

A series of announcements line up to a big theme: Services are the next big opportunity.

[hugging face]

Adding Benchmaxxer Repellant to the Open ASR Leaderboard

[openai]

Uber uses OpenAI to help people earn smarter and book faster

Uber uses OpenAI to power AI assistants and voice features that help drivers earn smarter and riders book faster across a global real-time marketplace.

[openai]

Singular Bank helps bankers move fast with ChatGPT and Codex

Singular Bank built Singularity, an internal assistant using ChatGPT and Codex to help bankers save 60–90 minutes daily on meeting prep, portfolio analysis,…

[openai]

How frontier firms are pulling ahead

OpenAI’s B2B Signals research shows how frontier enterprises deepen AI adoption, scale Codex-powered agentic workflows, and build durable competitive advantage.

[openai]

Introducing ChatGPT Futures: Class of 2026

Meet the ChatGPT Futures Class of 2026—26 student innovators using AI to build, research, and drive real-world impact. Discover how this generation is…

[vercel]

Auto-add Git committers to your team

Pro teams can now choose how Git committers to private repositories are added to their Vercel team. Choose your approval preference in .team settings Learn…

[simon willison]

datasette-referrer-policy 0.1

Release: datasette-referrer-policy 0.1 The OpenStreetMap tiles on the Datasette global-power-plants demo weren't displaying correctly. This turned out to be…

[simon willison]

Our AI started a cafe in Stockholm

Our AI started a cafe in Stockholm Andon Labs previously started an AI-run retail store in San Francisco. Now they're running a similar experiment in…

[latent space]

🔬Doing Vibe Physics — Alex Lupsasca, OpenAI

The full story of how GPT‑5.x derived new results in theoretical physics and quantum gravity.

[openai]

Unlocking large scale AI training networks with MRC (Multipath Reliable Connection)

OpenAI introduces MRC (Multipath Reliable Connection), a new supercomputer networking protocol released via OCP to improve resilience and performance in…

[openai]

GPT-5.5 Instant System Card

[openai]

GPT-5.5 Instant: smarter, clearer, and more personalized

GPT-5.5 Instant updates ChatGPT’s default model with smarter, more accurate answers, reduced hallucinations, and improved personalization controls.

[vercel]

Query observability metrics using the Vercel CLI

You can now access Observability Plus metrics in the Vercel CLI. Query observability data for any Vercel team or project using the new command. Coding agents…

[vercel]

How KIKO Milano scales for Black Friday

KIKO Milano on Vercel: Environments without the bottleneck Shipping faster without the operational tax No more performance variability during traffic spikes…

[simon willison]

datasette-llm 0.1a7

Release: datasette-llm 0.1a7 Mechanism for configuring default options for specific models. Part of Datasette's evolving support mechanism for plugins that use…

[simon willison]

llm-echo 0.5a0

Release: llm-echo 0.5a0 New -o thinking 1 option to help test against LLM 0.32a0 and higher. This plugin provides a fake model called "echo" for LLM which…

[openai]

New ways to buy ChatGPT ads

OpenAI expands ChatGPT ads with a beta self-serve Ads Manager, CPC bidding, and enhanced measurement tools—built to protect privacy and keep conversations…

[latent space]

[AINews] The Other vs The Utility

a quiet day lets us reflect on the nature of AI "character" in the Clippy vs Anton debate

[android devs]

Gemini and Firebase AI Logic enabled Karrot to increase sales with a translation feature built in under 2 weeks

Posted by Thomas Ezan, Sr Developer Relations Engineer and Tracy Agyemang, Product Marketing Manager Karrot is a hyperlocal, community-driven peer-to-peer…

[vercel]

How General Intelligence used agents to build an agent platform on Vercel

General Intelligence on Vercel Migrating Cofounder's Python backend to Vercel Running Cofounder as a multi-tenant app on Vercel 8-person team (5 engineers)…

[vercel]

Introducing deepsec: The security harness for finding vulnerabilities in your codebase

Today we’re open sourcing : a security harness powered by coding agents. It runs on your own infrastructure and surfaces hard-to-find issues in large…

[latent space]

[AINews] AI Engineer World's Fair — Autoresearch, Memory, World Models, Tokenmaxxing, Agentic Commerce, and Vertical AI Call for Speakers

a quiet day lets us make a call for speakers!

[latent space]

[AINews] Agents for Everything Else: Codex for Knowledge Work, Claude for Creative Work

a quiet day lets us reflect on coding agents "breaking containment"

[vercel]

How GitBook serves 30,000 sites with sub-second content updates

GitBook on Vercel 30,000 documentation sites hosted on a single Vercel deployment 120 million monthly page views served from the edge 40,000 cache…

[vercel]

Postgres connections now work through Sandbox firewall

can now connect to hosted Postgres databases, including , , , , and . To enable a connection, add the database host to your Sandbox's allowed domains.Vercel…

[github blog]

GitHub Copilot CLI for Beginners: Interactive v. non-interactive mode

Learn the difference between CLI interactive v. non-interactive modes. The post GitHub Copilot CLI for Beginners: Interactive v. non-interactive mode appeared…

[vercel]

Grok 4.3 on AI Gateway

Grok 4.3 is now available on . The model has a 1M token context window and improvements in accuracy, tool calling, and instruction following.Vercel AI Gateway…

[vercel]

Custom tags available in beta on Vercel Sandbox

As teams scale isolated environments for AI agents, code generation, or dev workflows, keeping track of which sandbox belongs to whom, and why, becomes…

[latent space]

[AINews] The Inference Inflection

a quiet day lets us reflect on the growing implications of the inference age

[hugging face]

Granite 4.1 LLMs: How They’re Built

[latent space]

[AINews] not much happened today

a quiet day.

[hugging face]

DeepInfra on Hugging Face Inference Providers 🔥

[vercel]

Vercel now supports Pro plan in Stripe Projects

You can now sign up for or upgrade to a Vercel Pro plan directly from Stripe Projects using (SPTs). Agents and developers can manage plan changes…

[hugging face]

Introducing NVIDIA Nemotron 3 Nano Omni: Long-Context Multimodal Intelligence for Documents, Audio and Video Agents

[latent space]

[AINews] ImageGen is on the Path to AGI

reflecting on the continued GPT-Image-2 explosion

[latent space]

Physical AI that Moves the World — Qasar Younis & Peter Ludwig, Applied Intuition

Applied Intuition puts the AI in mining rigs, drones, trucks, warships and physical vehicles in the most adversarial environments imaginable. We dive in with…

[apple developer]

Now Available: Monthly Subscriptions with a 12-Month Commitment

Today, we’re introducing a new way that people can pay for your auto-renewable subscriptions on the App Store: monthly subscriptions with a 12-month…

[hugging face]

How to build scalable web apps with OpenAI's Privacy Filter

[latent space]

[AINews] DeepSeek V4 Pro (1.6T-A49B) and Flash (284B-A13B), Base and Instruct — runnable on Huawei Ascend chips

The prodigal Tiger returns... but is no longer the benchmarks leader.

[apple developer]

Lykke Studios: In pursuit of puffy perfection

The delightful game puffies. combines the satisfying snap of a jigsaw puzzle with the nostalgic delight of a sticker book. This 2025 Apple Design Award…

[karpathy]

microgpt

.post-header h1 { font-size: 35px; } .post pre, .post code { background-color: #fcfcfc; font-size: 13px; /* make code smaller for this post... */ } This is a…

[lil'log]

Why We Think

Special thanks to John Schulman for a lot of super valuable feedback and direct edits on this post. Test time compute (Graves et al. 2016, Ling, et al. 2017,…

[lil'log]

Reward Hacking in Reinforcement Learning

Reward hacking occurs when a reinforcement learning (RL) agent exploits flaws or ambiguities in the reward function to achieve high rewards, without genuinely…

[karpathy]

Deep Neural Nets: 33 years ago and 33 years from now

.post-header h1 { font-size: 35px; } .post pre, .post code { background-color: #fcfcfc; font-size: 13px; /* make code smaller for this post... */ } The Yann…