
Palo Alto Networks buys Applied Compute for $4.5B as earnings beat
The security company is paying in stock for a startup that trains and runs open-weight models, and its quarterly revenue rose 39% to $3.44 billion.
762 stories

The security company is paying in stock for a startup that trains and runs open-weight models, and its quarterly revenue rose 39% to $3.44 billion.

Anthropic confirmed the price above its $70 to $78 range and sold 980 million shares, about $81.3 billion, valuing it near $1.5 trillion.

The mid-sized Gemini 4 model targets coding agents and office work, and Google says it finishes tasks for 32% less than Carbon.

Cursor and Codex start at $8 a month with free tiers, Claude Code at $20, and only Claude Code and Codex appear on Terminal-Bench 4.1.

An energy-based model rates how well an answer fits its input and searches for the best fit. A new paper uses the idea to solve 99.9% of hard Sudokus with a 1.8M-parameter model.

Google rebuilt the app so Gemini Live, camera help and Nano Banana images work on entry-level Android (Go edition) devices.

CommSec led a local retail offer that closed on November 17. Trading on Nasdaq is expected to start in the early hours of Friday, November 20, Sydney time.

SpaceX priced at $135 in June and valued itself at about $1.765 trillion. Anthropic's $70 to $78 range implies about $916 billion to $1.02 trillion, before any upsizing.

Past AI Mode searches are saved under Search Services History, and Google's new temporary threads keep a conversation out of your account entirely.
A California report recommends state-set standards, outside evaluators, emergency shutdown orders and a cap on self-improving AI, but no bill or penalties yet.

F5 shipped nginx 1.30.7 and 1.31.9 on November 18. The worst flaw dates to 2015 and needs only HTTP/2 turned on.
Google's world model adds sound and mid-session prompts, Genie 3 sessions grow to 3 minutes, and Pro users get a waitlist.

Both chips reach 128GB of unified memory, but Apple lists 614GB/s of bandwidth while NVIDIA quotes 1 FP4 petaflop and runs CUDA on Windows on Arm.
The Bengaluru startup's Series A mixes $26 million in equity led by Accel with $21 million in venture debt.
The app runs coding tasks locally or in a cloud sandbox, several at once, with Linux support still to come.

Your HF token still works after chat models on Inference Providers moved to OpenRouter's routing on November 16, but your own provider keys no longer do for language models.
Both start at $20 a month, but Devin is a cloud agent that runs models from several labs, while Claude Code starts in your terminal and runs Claude only.
Language models predict the next token and diffusion models turn noise into pictures. Here are the papers behind them, the products built on them and where they fail.
OpenAI shipped GPT-6.2 Astra and Sol on November 17, but its cheapest model is still GPT-6 Luna, at $0.10 per million input tokens.

The investigation reportedly looks at Nvidia's chip sales in Asia and its decision to stop breaking out Singapore revenue.

Nvidia's Arm-based chip now ships in compact Windows desktops with up to 128GB of unified memory, enough to run 120-billion-parameter models locally.

There is no plain download button: install over USB, publish to your personal Play library, or export the ZIP and build the APK in Android Studio.

Diffusion models learn to undo noise step by step. The idea powers Stable Diffusion and is now being tried for language in models like LLaDA and Gemini Diffusion.

Samsung lists camera, health, accessibility and App Lock changes but no new AI features, and gives no date for the stable release.