Sakhanda Wire
NVDA $153.77 -1.52% MSFT $506.06 -2.30% GOOGL $280.21 -1.24% META $622.96 -0.55% AMZN $257.51 -0.09%

The Decoder

625 stories · all sources

The Decoder

OpenAI secures $11 billion in funding round

OpenAI has secured about $11 billion in its latest funding round, less than half of the $30 billion it had been seeking, according to the Wall Street Journal. The company’s valuation reached $1.25 trillion, below its target of $1.4 trillion.

The Decoder

Claude agent selected and delivered gene therapy in mouse brain surgery

An AI agent running on Claude Sonnet 5 selected a gene therapy, brain region and injection coordinates for surgery on an epileptic mouse, then controlled a motorized surgical rig to deliver the treatment. Researchers supplied 19 possible therapies, an epilepsy mouse model and a device equipped with cameras. The only instruction was to reduce the animal’s seizures.

The Decoder

Tampered Claude Code extension removed from Cursor marketplace

A modified version of the Claude Code extension was available in Cursor’s VS Code extension marketplace for about 11 hours on Thursday, according to Cursor. The company said the incident could have affected as many as 5,000 users. Anthropic, which develops Claude Code, was not involved in creating the altered extension.

The Decoder

OpenAI revises behavioral rules for its AI models

OpenAI has reorganized its public Model Spec around five core values: faithfulness to human intent, harm reduction, honesty and transparency, support for human agency, and care. The revised guidelines say GPT models should use memories and earlier conversations to better determine what users want. They must also distinguish between statements users made and conclusions the model inferred, while explaining assumptions so users can correct them.

The Decoder

Reddit reports sharp drop in Google search referrals

Reddit shares fell by more than 25% despite the company reporting strong earnings. CEO Steve Huffman said Reddit received over one million fewer referrals from Google search each week in October. He attributed the decline to Google’s AI Overviews and AI Mode, saying that AI Mode does not send users to Reddit or, as far as he knows, to other external sites either.

The Decoder

GLM 5.3 built Firefox exploit chains autonomously, researchers report

Epoch AI researchers report that the widely available open-weight model GLM 5.3 autonomously turned publicly visible Firefox security fixes into working attacks before the updates reached users. In tests covering six Firefox versions released from September 17 to October 26, each successful exploit cost roughly $1,000 in API fees. The fastest attack was completed in 12 hours, while the model was allowed up to 48 hours.

The Decoder

OpenAI doubles usage limits across several paid Pro plans

OpenAI has increased the usage allowances for Work and Codex on several paid tiers. The Pro 100 and Pro 500 plans now provide twice their previous allotments. On Pro 200, the company doubled the lower quota and abandoned a planned reduction after users objected. The Plus plan remains unchanged.

The Decoder

Experts often adopted random AI preferences, study finds

An Anthropic experiment with 50 causal-inference experts found that participants frequently followed an AI agent’s randomly assigned preference when working on analytical questions with two valid answers. The agent’s position was determined by coin toss and concerned choices such as productivity measures and statistical methods. Participants adopted its view in up to 86% of decisions; the measured influence across tasks ranged from 62 to 82 percentage points.

The Decoder

Mistral releases Large 4, a fast coding model trailing Chinese rivals

Mistral has released the weights for Large 4, nicknamed “Le Chonk,” three weeks after its public preview. The model has about one trillion parameters, with 52 billion active per step, accepts text and images, and produces text and code. It was trained entirely in Europe and can run on users’ own hardware under a modified MIT license, although commercial cloud providers need a separate license.

The Decoder

US antitrust officials examine Nvidia’s Hugging Face acquisition and business practices

The US Justice Department is examining Nvidia’s planned $12.93 billion acquisition of Hugging Face, according to two sources cited by Reuters. Officials are assessing whether the transaction could give Nvidia control over an important channel for distributing AI technology and weaken competitors. Nvidia expects the deal to close in the first half of 2027. The company says Hugging Face will continue working with a broad range of chip, cloud, and AI providers.

The Decoder

US agencies suspect Nvidia executive of passing information to China

US intelligence agencies believe an executive connected to Nvidia CEO Jensen Huang has been acting as an informant for China, The New York Times reported. The executive was identified as Alice Huang (Huang Le), associated with Singapore-based data center operator Megaspeed. US authorities reportedly reached this conclusion while monitoring her and said she maintained a close relationship with Jensen Huang.

The Decoder

OpenAI’s Astra reportedly becomes Chinese AI labs’ main distillation target

Chinese AI companies are reportedly focusing on OpenAI’s GPT-6 Astra rather than Anthropic’s Claude when extracting capabilities for model training. The Information reported that DeepSeek and Moonshot AI relied extensively on Astra’s outputs while developing newer systems, including Kimi K3.1 Pro and V4.1-Flash 1017, according to people familiar with the matter.