AI News Archive - Browse Page 132 of 282
Browse AI news articles covering LLMs, tools, research, and industry trends
Lean4Agent launches FormalAgentLib to model and verify workflow consistency
Every agent workflow fails eventually. Usually in a weird, unpredictable way you can't debug.
Python Multi‑Agent System Built via OOP Class Blueprint for Agents
You don’t need a sprawling framework to build a multi‑agent system. You need a blueprint.
Perplexity's Search as Code lets AI build pipelines, improving performance
Every AI company selling search is secretly in the infrastructure business. They just don't want to admit it.
Study Finds No One-Size-Fits-All Strategy for Multi-Agent Communication
Forget finding the perfect way for AI agents to talk. There isn't one. A new study tested five standard communication strategies across different...
Reddit releases AI comment archive to study LLM persuasion tactics
Reddit conducted a quiet, unsettling experiment. For a period, the platform allowed moderators to secretly flood specific forums with comments...
xAI used Anthropic’s Claude via personal accounts after access revoked for months
Elon Musk’s xAI trained its models by borrowing from a rival long after the rival told it to stop. The technical term is data mooching.
Meta launches Hatch AI agent, its first paid product, priced up to USD 200/month
Two hundred dollars a month. That’s the price of a decent car payment, or a generous grocery run.
SpaceX inks USD 920 M/month deal with Google for 110,000 Nvidia AI chips
SpaceX is renting out its computers for a billion dollars a month. The client is Google, which happens to own a chunk of SpaceX and desperately needs...
Open‑source voice model listens continuously, decides to speak every 0.4 seconds
Most voice assistants are terrible listeners. They wait for a magic word, ignore the noise in between, and treat a conversation like a series of...
S&P 500 refuses entry for SpaceX, OpenAI and Anthropic, analysts surprised
The S&P 500 just drew a line in the sand. And it’s not the one Elon Musk wanted. On June 4, S&P Dow Jones Indices refused to bend its own rules for...
On‑Policy vs. Off‑Policy: TD Learning Updates Using Next‑State Estimates
Reinforcement learning has a central, simple tension: do you learn from what you did, or what you could have done? TD learning forces the question.
NSA uses Anthropic's Mythos AI model for offensive cyber ops against China, Iran
The National Security Agency is hacking foreign governments with a corporate AI. It’s adapting Anthropic’s Mythos model for cyber attacks on China...
Study examines temporal preference concepts in large language models
AI models are making our choices now. That means they're constantly facing a very human dilemma: grab the immediate reward, or wait for a better one...
Errorquake-10k Benchmark Scores 10,000 LLM Responses on 0-4 Severity Scale
Accuracy alone is a lie. It flattens every mistake into a single number, hiding the gulf between a model misremembering a date and one concocting a...
Nvidia plans PC reboot, Apple unveils smart glasses on Vergecast
The laptop you bought last year is obsolete. Nvidia, the trillion-dollar chip architect for data centers, is now redesigning the personal computer.
Open LLM v2, 12‑benchmark suite, LiveBench show d_eff 2.86‑4.80
The numbers are deceptively tight. Three independent leaderboards, Open LLM v2, a twelve-benchmark suite, LiveBench, all converge on a narrow band of...
TensorFlow Emotion Dataset with 54,263 Texts Shows Class Imbalance
54,263 labeled texts. That’s the raw size of the TensorFlow Emotion Dataset, a resource released under Apache 2.0 for anyone building emotion...
Three SpaCy Tricks Speed Up Production-Grade Text Processing
Processing text at scale doesn't have to be slow. Most spaCy pipelines handle documents individually, a method that wastes CPU cycles and complicates...
Gemini 3.5 adds action‑taking to run complex multi‑step workflows in apps
Every AI assistant so far has just been a chatbot with better search. You ask, it answers. That's the old game.
Zhipu AI employs Muon Optimizer and Muon Split in GLM-4.5 and GLM-5 pretraining
The wall isn't made of silicon. At Zhipu AI, engineers hit it while training GLM-4.5.