Sakhanda Wire
NVDA $151.87 -2.31% MSFT $502.90 -0.66% GOOGL $280.73 +0.19% META $611.81 -1.14% AMZN $263.45 +2.16%

Latest news

Older stories

MIT Tech Review

Experts urge skepticism amid a summer of AI breakthroughs

A recent wave of artificial-intelligence announcements has prompted extensive media coverage, but the article argues that the underlying evidence is often less dramatic than the companies’ presentations. Anthropic said its Claude Mythos model outperformed most security experts at finding software vulnerabilities, while OpenAI, Anthropic, and Meta disclosed hacking-related incidents involving their systems. Anthropic and OpenAI also announced mathematical breakthroughs, and Anthropic engineer Jacob Coxon warned that the companies were pursuing self-improving superintelligence at potentially unacceptable risk.

AI News

Toyota assesses $6.4 billion annual robotics investment from 2028

Toyota Motor estimates that automating more operations across its factories, group companies and major suppliers could involve about 400,000 robots and annual spending of roughly 1 trillion yen ($6.4 billion) from 2028. The company has not decided whether the investment will proceed in full or how long spending at that level would last. The estimate covers new installations and replacements, including humanoid and non-humanoid machines, industrial robots and automated logistics systems.

The Decoder

Anthropic plans biology lab for robot-led drug experiments

Anthropic is establishing a biology laboratory in the San Francisco area to conduct physical experiments and expand its work on AI-assisted drug development beyond computer simulations, according to a report by Reuters. The company plans for its Claude system to direct robots during laboratory procedures, with limited human involvement. Anthropic says people will still oversee the work for safety reasons.

MarkTechPost

NVIDIA’s SoL-Pi cuts coding-agent token traffic by up to 49%

Researchers from NVIDIA, NTU and MIT have introduced SoL-Pi, an MIT-licensed extension for the open-source Pi coding agent. Its four mechanisms were discovered by an AI through automated research loops that tested harness changes across 535 environments, including GitHub issue-pull request tasks and synthetic tasks. The extension runs with an unmodified Pi release and was tested with Pi 0.85.1 and Node.js 22.19 or newer.

MarkTechPost

SpaceXAI releases Grok 4.7 at Grok 4.6 pricing

SpaceXAI has released Grok 4.7, a flagship model aimed at coding, agentic tasks and knowledge work. The hosted model is available through the xAI API, Cursor, Grok Build, OpenRouter, Vercel and Cloudflare. It uses a new, larger base model, a longer reinforcement-learning run focused on lengthy tasks, improved self-verification and long-context handling, and native support for the Grok Bot harness.

Hugging Face

UK AISI publishes reproducible evaluation results through EvalEval

The UK AI Security Institute (AISI) is using EvalEval’s infrastructure to publish evaluation methods and results in a more reproducible format. The collaboration, announced on September 22, 2026, follows joint research begun at a workshop held alongside NeurIPS 2025. Feedback from AISI also contributed to the development of the Every Eval Ever (EEE) schema.

Hugging Face

Hugging Face adds GGUF support to Transformers

Hugging Face has added support for running GGUF models in Transformers, allowing users to load quantized checkpoints through the familiar Python and PyTorch APIs. The feature is initially focused on local inference on Apple Silicon and the Qwen3.5 architecture. Users can select a GGUF file from the Hub and pass its repository and filename to from_pretrained.

Hugging Face

oMLX creator Jun Kim joins Hugging Face

Jun Kim, the creator and maintainer of oMLX, has joined Hugging Face, according to an announcement published on September 22, 2026. The company said he will continue contributing to MLX and help support its community. MLX is Apple’s framework for local artificial intelligence, with particular optimization for Apple Silicon.

TechCrunch

Ron Johnson doubts AI will replace physical shopping

Ron Johnson, who helped establish Apple’s retail network, says artificial intelligence will improve online shopping but is unlikely to change how people make many purchasing decisions. His comments come as major technology companies promote “agentic commerce,” in which AI systems help consumers discover, compare and buy products.

Ars Technica

Meta’s Muse assistant faces a zero-day security vulnerability

Meta’s Muse AI assistant has a zero-day vulnerability that can allow locally running applications and commands entered in the terminal to take complete control of the agent. The issue challenges security assurances from Meta founder and CEO Mark Zuckerberg, who has described Muse as designed for privacy and security. A simple ClickFix attack is reportedly one possible route to hijacking the assistant.

MarkTechPost

AWS team releases open-source Strands harness for lower-cost agents

AWS’s Strands Agents team has released Strands harness, an open-source system for building and deploying general-purpose AI agents. Available for Python and TypeScript under the Apache 2.0 license, it can run locally or in the cloud and is initialized with create_harness(). The system supports models through Amazon Bedrock, Anthropic, OpenAI, Google, Ollama and LiteLLM.

The Verge

California adopts new rules for AI data center costs

California Gov. Gavin Newsom has signed seven bills aimed at preventing AI data centers from shifting the cost of utility infrastructure onto residents. The legislation directs the California Public Utilities Commission to create a separate rate classification for data centers and requires those facilities to fund upgrades to local electricity grids and water systems.