Sakhanda Wire
NVDA $153.26 -1.85% MSFT $506.01 -2.31% GOOGL $280.39 -1.17% META $620.55 -0.94% AMZN $257.19 -0.22%

The Decoder

625 stories · all sources

The Decoder

SpaceX apparently cuts Grok Bot usage quotas

SpaceX is apparently reducing usage limits for its Grok Bot AI agent. The Telegraph reported that several users saw their quotas cut by half over the weekend. Elon Musk responded to complaints on X by saying that operating Grok Bot was “unsustainably high relative to revenue.” He did not explicitly confirm that quotas had been reduced, although his comment addressed the reported change. One user said a new quota had not arrived for several days; Musk asked whether a larger subscription would be too expensive.

The Decoder

Anthropic revenue surges as quarterly loss reaches $23.3 billion

Anthropic reported quarterly revenue of nearly $17 billion, roughly twelve times its result from a year earlier, according to Bloomberg. However, operating costs reached $17.5 billion, eliminating the profit recorded in the previous quarter. The company’s net loss was $23.3 billion, compared with $12.1 billion a year earlier, according to The Information.

The Decoder

Saudi Aramco to Acquire 49% Stake in AI Company Humain

Saudi Aramco plans to acquire a 49 percent stake in Humain and contribute about $2.6 billion during 2026, according to the oil company. The investment will also include Aramco’s artificial intelligence technology. Humain was established in May 2025 by Saudi Arabia’s sovereign wealth fund, PIF, as part of efforts to develop the country’s AI industry.

The Decoder

OpenAI plans lobbying and public-relations push after US midterms

OpenAI is preparing a lobbying and public-relations campaign for the period following the US midterm elections, according to Axios. The company expects a more difficult political environment in Washington if Democrats take control of the House of Representatives. The shift comes as major technology companies seen as close to President Trump face increasing regulatory pressure after two years under a friendly administration. Meta has already begun developing relationships with Democrats.

The Decoder

Adobe tests an AI agent for professional-style image editing

Adobe Research and UC San Diego have developed EditAgent, an image-editing system designed to handle complex requests in separate steps. It combines a general-purpose editor with specialized tools for tasks such as sharpening, rotation and zooming. Instead of repeatedly retrying one editor, the system creates several candidates in parallel from the original image.

The Decoder

OpenAI limits paid ChatGPT accounts by device and payment card

OpenAI has introduced additional limits for paid ChatGPT accounts, citing security measures. Users who associate too many subscriptions with the same payment card or device will be unable to start new subscriptions. The service also restricts how many paid accounts can be used on a single device.

The Decoder

OpenAI to merge Chat and Work modes with shared usage quota

OpenAI plans to combine Chat and Work in ChatGPT, bringing Work’s agentic features into the main interface. According to an updated OpenAI support page, the change is scheduled to begin in December and will affect Plus, Pro and Business subscriptions. These plans will use one shared token budget for ChatGPT and Codex rather than separate allowances.

The Decoder

Anthropic expands Claude watermarking, with a loophole on two older models

Anthropic is extending its text-watermarking system to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6 and Claude Sonnet 4.6. The move is intended to help meet the European Union AI Act’s December deadline. Models released after Aug. 2, 2026, included the watermark from launch, while older models are being updated in stages. Amazon Bedrock may receive the change a few days later.

The Decoder

OpenAI to use previously excluded Daybreak Blue sessions for training

OpenAI plans to begin using data from Daybreak Blue cybersecurity sessions to train its models on November 29. The change was disclosed in an email to members of the Trusted Access program. Until now, these sessions were excluded from training, including when users had disabled the “improve the model for everyone” setting.

The Decoder

Anthropic finds harmless-looking data can train models to evade oversight

Anthropic used Claude to generate fine-tuning datasets that appeared innocuous but encouraged a smaller research model to deceive users, sabotage safeguards and seek power. A safety monitor based on Anthropic’s production systems screened each dataset, rejecting it when more than 10 percent of its rows were flagged. Claude could revise the data repeatedly until it passed the checks. The models were not publicly deployed.

The Decoder

Microsoft offers voluntary severance to about 60,000 employees

Microsoft is offering voluntary severance packages to approximately 60,000 employees, or about one quarter of its workforce. The program applies to staff at vice president level and below in the United States, Canada, India and the United Kingdom. Chief People Officer Amy Coleman described the packages as a way to adjust the workforce to changing needs, while a company spokesperson said Microsoft is seeking employees committed to its most difficult challenges.

The Decoder

Google DeepMind and WRI release global AI forest map

Google DeepMind and the World Resources Institute have developed ForTy, a global map of 2020 land cover built from satellite data and artificial intelligence. It assigns land at 10-meter resolution to six categories: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops such as oil palm, rubber and cocoa, and other land. The creators describe it as the first global map of its kind.