Old-school classifiers are still generally better than Jev [P]
There’s a lot of hype right now around decision models like Jev. But in most real use cases, you have some labeled data, and can…
AI · Tools · Online creation
Latest AI news
Every AI story worth reading, collected hourly from the labs, the tech press, research and the creator economy. Sorted by time, grouped by topic.
3,455 stories · Every headline links to the original publisher. HapnGo curates and groups; we do not copy articles.
US antitrust regulators are scrutinizing Nvidia's Hugging Face deal and sales practices
AI summary The US Justice Department is reviewing whether Nvidia's planned $12.93 billion purchase of Hugging Face could hand it control of a central AI distribution platform and harm rivals, according to two sources cited by Reuters. Nvidia expects the acquisition to be completed in the first half of 2027. A broader antitrust probe, begun under Joe Biden and continued under Donald Trump, covers Nvidia's sales tactics, its Groq licensing arrangement and its takeovers of Run:ai and SchedMD. In France, the competition authority is separately investigating the chipmaker over reliance on its CUDA software and its stakes in cloud firms such as CoreWeave.
Quoting Ben Hylak
every product i use should just be a headless mcp, that feeds into an agent, where i never have to think about anything…
Mistral Says Its AI Model ‘Le Chonk’ Has Leapfrogged China in the Open-Weight Race
The French AI lab claims its latest release matches or outperforms even the best models from DeepSeek and Z.ai in coding and…
Daily Search Forum Recap: October 29, 2026
AI summary Google is bringing ads in AI Mode to ten additional countries, among them Canada, India and Australia, with Brazil, Mexico and the United Kingdom to follow in Q4, and has started multiturn ads in the US. Its Q3 2026 results showed search ad revenue rising 10.5% to $82 billion and total revenue climbing 25% to $127.6 billion. Microsoft said its search and advertising revenue grew 12% in its first quarter of fiscal 2027. Google is also reportedly testing the removal of the source citations card from desktop AI Overviews and has opened a preview of its Connected Apps Platform for Search and Gemini.
Mistral releases weights for its trillion-parameter AI model, targeting cybersecurity
Mistral says its new model can afford organizations a kind of control over its cyber defenses that closed models can't.
Anthropic is asking Wall Street to believe in AGI
Today on Decoder, I’m joined by Joey Brookhart, an analyst at the chip and AI research firm SemiAnalysis, to help me dig into the…
A one-person studio shipped a Steam demo of a 12-chapter FPS coded by AI agents in 20 working days. Here is what the agents could and could not do.
I am Ekin, a one-person game studio in Turkey. On 28 October the free demo of Iron Curtain Call went live on Steam: a fast…
Chinese spies targeted Nvidia CEO Jensen Huang, according to The New York Times
AI summary US intelligence believes Alice Huang (Huang Le) of Singapore data center firm Megaspeed, who is close to Nvidia chief Jensen Huang, feeds information to Beijing, per The New York Times. Trump was told roughly a year ago but took no action, and the Nvidia CEO was never warned. Washington also suspects Megaspeed helps China obtain Nvidia chips; a Malaysian affiliate bought at least $4.8 billion in AI hardware from June 2024 to February 2026. Nvidia says Megaspeed gets no sensitive information, Megaspeed has opened an internal probe, and China's embassy rejected the claims.
Hybrid AI Patterns: Thirteen Ways to Couple Language Models with Rules, and How to Choose Among Them
Language models handle text well, and rule engines give conclusions with proofs. Systems that combine the two are usually…
Granite Switch 4.1.1 — Update models like you update software
AI summary IBM has released Granite Switch 4.1.1, a patch that stays fully compatible with version 4.1, which came out just over a week earlier, and requires no migration. The update brings a refreshed Granite Speech ASR model that transcribes 10 languages with built-in punctuation and capitalization, plus a revised Decision adapter in granitelib-core that scores better on the Jev Decision Index test set. Only the new adapters were trained, leaving the base model unchanged. IBM says it is preparing tutorials and an Adapter SDK so outside developers can create and version their own adapters.
Cheap Retrieval and Observation Credit Assignment from One Quantity: The Value of Information
Memory in reinforcement learning raises two problems. Attention finds the relevant item by scoring the whole store, which is…
HEMO: Harness Self-Evolution for Multi-Objective Combinatorial Optimization
Automatic operator design (AOD) has increasingly adopted large language models (LLMs) to generate problem-specific search and…
Anthropic’s business plan: Build God or die trying
The Anthropic S-1 seems to have provided the financial justification for the company to spend as much as it possibly can on…
FIDE: Fiber-Informed Differentiable Experimental Design for Reference-Relative Law Dynamics
Aggregate measurements generally leave an evolving population law underdetermined, so different measurement systems can be…
Human--AI Interactions as First-Class Data: A Formal Interaction Model and Runtime for Explainable AI Collaboration
AI assistants are increasingly embedded in computational notebooks, scientific workflows, software engineering tools, and…
Efficient Learning of Mechanism Designs and Compositionality
Learning mechanism behaviour requires a representation of how connected parts jointly constrain motion. We study whether exposing…
How Much Should a VLA Trust Its Reasoning? Step-Level Guidance as a Measured Action-Value Function
Reasoning vision-language-action (VLA) policies write a reasoning trace and then decode the action from it. A wrong trace yields…
OpenAI's Astra becomes the main target for Chinese AI labs after Anthropic tightens its defenses
AI summary DeepSeek and Moonshot AI reportedly relied heavily on outputs from OpenAI's GPT-6 Astra to train recent models, including V4.1-Flash 1017 and Kimi K3.1 Pro, according to people familiar with the matter cited by The Information. Such distillation requires coaxing models into exposing hidden reasoning, which both OpenAI and Anthropic prohibit. Anthropic's Claude was previously the favored target, but the company says new safeguards reduce extractable data by 90 to 99 percent. OpenAI, which flagged a coordinated campaign tied to Moonshot-linked people in late September, says attackers have adapted after each new defense.
Enhancing Atrial Fibrillation Detection Through Hyperparameter Optimization
This study explores the role of hyperparameter optimization (HPO) in enhancing the detection of Atrial Fibrillation (AF) using…
A physics-informed uncertainty-aware surrogate for $p$-$n$ junction diodes
We develop an uncertainty-aware physics-informed surrogate for a one-dimensional $p$-$n$ junction diode, conditioned on device…
Accumulating Pruning Constraints: New Exact Polynomial-Space Algorithms for Max 2-SAT
We present deterministic polynomial-space algorithms that solve Max 2-SAT on $n$ variables exactly in time $\Ostar(2^{(1-γ)n})$…
Does consciousness have to be biological, and what would it mean ethically if it doesn’t?
People argue that AI has no soul, that it isn’t conscious, that it cannot feel. I wonder how much of that certainty comes from…
Could Abdul El-Sayed’s plan for AI companies stave off catastrophe?
The Michigan Senate candidate is calling for public ownership and other reforms, offering a potential antidote to ideological…
Free, daily
The HapnGo Briefing, in your inbox
One email a day with the AI stories that matter, why they matter, and one tool worth trying. No spam, one-click unsubscribe.