Notes by Jin

AI

AI Daily — September 1, 2026

Models & Research Stress test shows cheaper AI evaluation methods can alter safety benchmark results — Vector Institute researchers ran three vision-language models on the BBQ and BBQ-V bias benchmarks under seven compute-saving conditions, each compared against a full-benchmark BF16 baseline. Larger batching held accuracy within 0.35 points and cut

AI

AI Daily — August 31, 2026

Industry & Funding ChatGPT's ad business hits $1 billion annualized run rate — ChatGPT Ads hits $1 billion annualised run rate under 200 days from launch — OpenAI reports tens of thousands of advertisers, availability in over 40 countries, and self-service buying via Ads Manager extending today to India, Europe,

AI

AI Daily — August 30, 2026

Models & Research 💡Astra from OpenAI and Fable 5.1 from Anthropic could be released within next week or two. fal launches H3 Max, a tuned version of MiniMax H3 — fal has released H3 Max, a post-trained variant of the open-weights MiniMax H3, developed by fal Research and served on

AI

AI Daily — August 29, 2026

Models & Research Anthropic paper: automated researchers beat human proposals on alignment benchmarks at 2.7% of the cost — Anthropic published "Automated Researchers Can Reliably Mitigate Alignment Failures" on August 28, led by fellow Chen Yueh-Han. Given ten benchmarks for specific misaligned behaviours, the automated systems improved every

AI

AI Daily — August 28, 2026

Tools & Open Source Google adds hotel checkout to AI Mode — and hands partners the liability — Google announced new capabilities for AI Mode in Search that let users track flight prices, view flight and hotel costs in points or miles, and book hotels without leaving the conversation — expanding the tool&

AI

AI Daily — August 27, 2026

Models & Research GLM-5.3-Flash (aka Ox Alpha): Frontier Intelligence at Flash Cost — Z.ai's launch post for GLM-5.3-Flash, the first natively multimodal model in the GLM-5 series — 320B total parameters, 18B active, using a hybrid linear + sparse attention architecture that Z.ai says cuts attention compute

AI

AI Daily — August 26, 2026

Models & Research AI models still struggle with classic intelligence puzzles — A review of long-standing puzzle and game benchmarks shows current AI systems still stumble on tasks designed to probe reasoning and general intelligence, offering a way for readers to test their own performance against the machines. MIT Tech Review

AI

AI Daily — August 25, 2026

Industry & Funding Hugging Face reportedly fielding $13B acquisition offers — Business Insider reported over the weekend, via TechCrunch, that Hugging Face has been approached at a valuation of $13 billion or more. No suitor is named, no deal has been reached, and the company has reportedly hired banks to evaluate

AI

AI Daily — August 24, 2026

Models & Research Why children still out-learn AI at language remains a mystery — Large language models consume roughly a hundred thousand times more words than a child hears while learning a first language, and researchers quoted by MIT Technology Review say they still can't explain the gap. Closing

AI

AI Daily — August 23, 2026

Models & Research OpenRouter Lists a Free 1M Context Stealth Coding Model Called Ox Alpha — OpenRouter is hosting "Ox Alpha," a stealth reasoning model listed on August 20, 2026 from an anonymous third-party provider. The listing describes it as built for coding, sustained agentic work, and production workloads,

AI

AI Daily — August 22, 2026

Tool & Open Source xAI Bundles Grok Bot Into More Paid Subscriptions — xAI has expanded access to Grok Bot, its agent product launched in beta on August 11, 2026. The bot is now included with SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, and Cursor Teams Standard and Premium. Shipped

AI

AI Daily — August 21, 2026

Industry & Funding AI training-data startup Micro1 scales rapidly amid data demand surge — Micro1 has hit a $500 million gross annualized revenue run rate as demand for data used to train AI models fuels fast expansion for the company and competitors in the space. TechCrunch ↗ My takeaway: Owning unique data

AI

AI Daily — August 20, 2026

Industry & Funding Stripe Agrees to Acquire OpenRouter — Stripe announced on August 19, 2026 that it has agreed to acquire OpenRouter, an AI model gateway that routes requests across what Stripe describes as more than 400 models from over 80 providers. Stripe frames the deal as an extension of its

AI

AI Daily — August 19, 2026

Tools & Open Source Cursor Launches Origin, Its Own Git Forge, in Early Beta — Cursor has released Origin, a git forge in early beta that hosts repos, mirrors from GitHub, supports standard git and pull requests, and lets teams browse code at cursor.com/codebase. It is limited to Pro,

AI

AI Daily — August 18, 2026

Models & Research Z.ai Ships GLM-5.3 on Post-Training Alone — Z.ai released GLM-5.3 on the same base model as GLM-5.2, with all gains attributed to scaled "post-training." The company reports a 50 percent coding improvement over GLM-5.2 on its in-house Z.ai Code

AI

AI Daily — August 17, 2026

Models & Research Researchers say encrypted reasoning blocks can be decoded through weaker models in the same API family — A new preprint reports an architectural flaw in how major providers handle hidden chain-of-thought. Because providers return reasoning as an encrypted block the client must send back on each turn, those

AI

AI Daily — August 16, 2026

Models & Research DeepMind ships sign-language-to-text dictation on Pixel 11 — Google DeepMind introduced SL2T, a sign-language-to-text translation model trained on more than 100,000 hours of data across more than 50 sign languages, with roughly a quarter of that data in American Sign Language. It powers sign-to-text dictation in Gboard

AI

AI Daily — August 15, 2026

Models & Research Meta Releases Muse Glimmer, a 30B Open Model Built for Always-On Local Agents — Meta has released Muse Glimmer, a 30 billion parameter open-weights model licensed under Apache 2.0 and sized to run on a single consumer GPU or a Mac. Meta positions it for long-running agent

AI

AI Daily — August 14, 2026

Models & Research Google Ships Gemini 3.7 Flash Three Weeks After 3.6 — Google released Gemini 3.7 Flash on August 13, 2026, pitched as its strongest workhorse model for coding and agents. Google claims gains over 3.6 Flash across coding, document, and workflow benchmarks, all self-reported. Introductory

AI

AI Daily — August 13, 2026

Models & Research xAI Releases Grok 4.6 With a Focus on Long-Running Agents — xAI has released Grok 4.6, a successor to Grok 4.5 built for multi-step agentic work and interactive or visual projects. xAI reports a score of 61 on the Artificial Analysis Intelligence Index, matching GPT-5.

AI

AI Daily — August 12, 2026

Models & Research Google tests AMIE on live video consultations in a simulated study — Google Research and Google DeepMind published what they describe as a first-of-its-kind study of AMIE, their experimental medical AI, holding real-time clinical video consultations. The randomized study used simulated consultations with patient actors and a group

AI

AI Daily — August 11, 2026

Models & Research Microsoft's MAI-Image-2.6 Debuts at No. 2 on the Arena Text-to-Image Leaderboard — Microsoft AI announced MAI-Image-2.6 on August 10, 2026, and says it ranked second on the Arena text-to-image leaderboard, ahead of models from Google, Meta, and xAI. Per Microsoft's own post,

AI

AI Daily — August 10, 2026

Models & Research Meta Open Sources Muse Glimmer, a 30B Agentic Model That Runs on One Consumer GPU —Meta Superintelligence Labs released Muse Glimmer, a 30-billion-parameter agentic model, under Apache 2.0 with weights live on Hugging Face. It was pre-trained on outputs from Muse Spark using logit distillation, then

AI

AI Daily — August 9, 2026

Model & Research OpenAI Tunes GPT-5.6 Sol for Everyday Chat and Opens Unlimited Text Chats to Free Users — OpenAI updated GPT-5.6 Sol in ChatGPT for Plus and Pro users, promising more focused answers and fewer factual errors, with a new slider to control reasoning effort. Free and Go

AI

AI Daily — August 8, 2026

Models & Research OpenAI Cannot Rule Out Critical Cyber Capabilities in Its Upcoming Astra Model — OpenAI says preliminary internal evaluations of Astra, an upcoming model, show significant gains in agentic coding and cybersecurity, enough that it cannot rule out "Critical" cyber capability under its Preparedness Framework. The company