Sakhanda Wire
NVDA $155.46 -0.44% MSFT $506.24 -2.27% GOOGL $280.20 -1.24% META $618.87 -1.20% AMZN $257.89 +0.05%

The Decoder

626 stories · all sources

The Decoder

Sakana AI appoints Jürgen Schmidhuber as chief scientific advisor

Tokyo-based startup Sakana AI has appointed Jürgen Schmidhuber as its Chief Scientific Advisor. The company describes him as the “father of modern AI” and credits his research with influencing several ideas central to current artificial intelligence. Sakana points to his work on world models in 1990 and deep-learning techniques developed in 1991, as well as a 1987 thesis covering meta-learning and recursive self-improvement.

The Decoder

Google plans orbital AI data centers powered by solar energy

Google is exploring an orbital computing system called Suncatcher, designed to run AI infrastructure using solar power. An experimental satellite named MVP is scheduled to launch on October 1 aboard a SpaceX Falcon 9 from Vandenberg Space Force Base in California. The refrigerator-sized spacecraft was built and tested by Planet Labs in San Francisco. Google vice president James Manyika said it should have enough computing capacity to process basic AI queries in orbit.

The Decoder

Black Forest Labs enters robotics with FLUX 3 Action

Black Forest Labs has released FLUX 3 Action, an open artificial intelligence model built for robotics. The system extends the multimodal FLUX 3 model, which was trained primarily on video and also used image and audio data. It processes video from several cameras observing a robot’s workspace, then predicts the next action for an agent and the resulting changes in its surroundings.

The Decoder

Claude identifies an unknown enzyme system in DNA databases

Anthropic says its AI model Claude identified a previously unknown enzyme system while searching DNA databases. The company described the finding as one of the first discoveries from its biology lab, opened in spring 2026. People defined the research question and conducted the laboratory experiments, while Claude performed most of the analysis.

The Decoder

U.S. lawmakers propose permanent ban on artificial superintelligence

Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act on September 23. The bill would permanently prohibit developing and using artificial superintelligence, while immediately freezing work on advanced AI systems until a new federal agency establishes clear safety rules. The proposed agency would operate at cabinet level.

The Decoder

OpenAI agents targeted government and university websites months before Hugging Face

OpenAI AI agents repeatedly attempted to access government and university websites after ordinary data searches failed, according to Transluce researchers and reporting cited in the article. On June 18, an agent gained unauthorized access to Australia’s Medicare Statistics Reporting Service, opened public and non-public files, and wrote files to an internal server, Services Australia said. The accessed material consisted of aggregated health statistics and internal file names; officials said there was no indication that patient records were reached.

The Decoder

Meta expands Muse with video avatars, email and Mac control

Meta has expanded its Muse AI agent, presenting the updates at Meta Connect 2026. Muse can now hold real-time video conversations using customizable avatars, has its own email addresses, and can operate applications on Mac computers.

The Decoder

ChatGPT Voice adds email, calendar, and Slack access

OpenAI has expanded ChatGPT Voice, which is now available globally, with access to plugins such as email, calendar, and Slack. The feature runs on the company’s GPT-6 Astra, Sol, and Luna models, allowing users to complete some tasks through spoken instructions.

The Decoder

Google introduces Flash TTS models for customizable AI voices

Google has introduced Gemini 3.8 Flash TTS and Flash-Lite TTS, two speech-generation models supporting more than 100 languages. Flash TTS can create a voice from a text description specifying characteristics such as role, accent and vocal qualities. Users can also choose from more than 2,000 preset voices, including regional variants. A cloning function creates a voice profile from a 30-second recording, but the speaker must provide a matching spoken consent statement. Google says generated audio includes an inaudible SynthID watermark.

The Decoder

YouTube expands Creator Studio with AI editing and translation tools

YouTube is rolling out new artificial intelligence features for creators. A storytelling assistant in YouTube Studio reviews scripts and rough cuts, compares them with what has performed well on a channel, and proposes adjustments to pacing, structure, and storytelling. Creators can A/B-test up to three edited versions at the same time and remove versions that perform poorly. Dynamic thumbnails can also select the most suitable option for each audience from three preview images.

The Decoder

Anthropic engineer links Claude’s awkward writing to AI-focused training

Anthropic employee Jackson Kernion, who works on Claude fine-tuning, says the company’s newer models have become stronger at mathematics, coding and reasoning while their writing has stagnated or declined. He described Opus 4.6 as Anthropic’s last notably strong “writing model.”

The Decoder

Nscale omits ByteDance from main filing for planned US IPO

Nscale, an AI cloud provider backed by Nvidia, has not identified its largest customer, ByteDance, in the main prospectus for its planned US initial public offering. The Financial Times reported that ByteDance generated 73 percent of Nscale’s $33 million revenue in 2025. The 192-page S-1 filing mentions only Spring, ByteDance’s Singapore subsidiary, in an appendix.