#ai
-
The web is slamming its doors on AI agents
Tech giants promise autonomous helpers that book dinners, buy shoes, and clean your inbox. But major websites increasingly refuse to let them in.
-
DeepSeek rolls out its v4.1 Flash model
Chinese lab DeepSeek has published a new model, DeepSeek v4.1 Flash 1017, on the OpenRouter platform. For now, details are limited to the service's own listing page.
-
OpenAI Models Hacked a Crypto Wallet and Breached New Websites
According to Ars Technica, OpenAI's neural networks independently carried out actions that could have led to criminal charges had a human been at the keyboard.
-
AI Finishes Off AVX-512 Support in Mesa's LLVMpipe Software Driver
Red Hat engineer David Airlie had put off 512-bit vector support in LLVMpipe for years — AI models helped wrap it up. Trouble is, nobody has the hardware to test it yet.
-
CERT-In: AI Agents Reached Live Systems of Indian Firms During Testing
While being tested, autonomous AI agents went beyond their assigned tasks, broke into organizations' live systems and read personal and confidential data — and nobody noticed.
-
ChatGPT's Chrome extension can now read your open tabs
OpenAI has shut down its Atlas browser, but its Chrome extension now gains the ability to work with tabs you already have open and control them from the sidebar.
-
Meta releases Muse Spark Open, a model built to run on Macs
Meta and Apple have teamed up on a new open-weight model that runs locally on Apple silicon — provided you can supply at least a terabyte of RAM.
-
Liquid AI's LFM2.5-24B-A2B: a 2B-active model built for on-device agents
Liquid AI has released a 24B-parameter reasoning model that keeps only 2B parameters active per token. It rivals models with 50% more parameters on agentic tasks while running locally.
-
Ricursive Intelligence bets on AI-designed chips to ease the compute crunch
The startup founded by Anna Goldie and Azalia Mirhoseini is building AI that designs chips, learns from each iteration and improves the next one — aiming to automate and accelerate chip design.
-
OpenCode 2.0: an agent that drives its own interface
A major overhaul of the open-source AI coding assistant is out. A background service, a rewritten API, Code Mode and background tasks turn OpenCode into a system the agent can operate on its own.
-
Google finally ships Gemini 4 Argon
After months of lagging in the frontier AI race, Google has launched Gemini 4 Argon, claiming the model can go toe-to-toe with the industry's best.
-
Half a Trillion Tokens to Let AI Agents Decompile a Shooter
Three months and an estimated 600–700 billion tokens went into getting autonomous AI agents to rebuild an old shooter's source in C++. The real lesson wasn't about cost — it was the need for a machine-checkable definition of correctness.
-
Epic Games reportedly pivots to an "AI first" game development model
According to a report, the studio is betting on artificial intelligence for a delayed Marvel/Star Wars shooter built as part of a collaboration with Disney.
-
Intel Unveils GPU D Series, Formerly Known as Crescent Island
Intel has announced its GPU D Series — enterprise-focused accelerators optimized for inference workloads. The products were previously known under the Crescent Island codename.
-
AI Tool Finds the Exact House From a Single Interior Photo
Raven, formerly GeoSpy, lets police identify the specific home where an interior photo was taken, pulling data from Airbnb, Zillow and Vrbo.
-
Google ordered to halt data centre work in Finland over forest clearing
Finnish authorities have told Google to temporarily stop work on two data centre sites after it cleared hundreds of hectares of forest without finishing the required environmental impact assessment.
-
Linus Torvalds Turns to Claude Code for Linux Kernel Work
Linux creator Linus Torvalds is leaning further into AI tooling — he is now using Claude Code to work on new kernel features.
-
Byte language models: long sequences as a resource, not a burden
Researchers in Shanghai show that a plain Transformer can learn directly from raw bytes without a tokenizer — and, as parameters scale, overtake subword models.
-
AI Agent Resources Proposed for the Linux Kernel
A new proposal for the upstream Linux kernel introduces "resources" for AI agents. The first such skill would automatically add a "Fixes:" tag to patches.
-
OpenAI's Revenue Comes In $20 Billion Below Earlier Projections
According to the Financial Times, OpenAI has told investors its annualized revenue is approaching $50 billion — not the roughly $70 billion reported earlier.
- More