Sakhanda Wire
NVDA $150.96 -2.89% MSFT $508.29 +0.40% GOOGL $282.44 +0.80% META $616.38 -0.40% AMZN $267.74 +3.82%

Latest news

Older stories

TechCrunch

Brad Lightcap to leave OpenAI and start a new venture

Brad Lightcap, one of OpenAI’s longest-serving executives, is leaving the company to “start something new.” In an internal message shared publicly on Tuesday, he described the decision as bittersweet and said he remained confident in OpenAI’s mission. Lightcap joined OpenAI in 2018 and spent four years as its chief financial officer. He became chief operating officer in 2022, holding that position until earlier this year, when an executive reorganization moved him to lead special projects.

TechCrunch

River AI raises $1.1 billion in round led by General Catalyst

River AI, an artificial intelligence startup founded by xAI co-founder Igor Babuschkin, has raised $1.1 billion in a seed/Series A round. General Catalyst and AMP PBC led the financing, which also included Nvidia, AMD Ventures, Y Combinator, and Temasek. AMP PBC was founded in 2026 by former Andreessen Horowitz general partner Anjney Midha.

The Decoder

Researchers report API flaw exposing AI reasoning traces

Security researchers led by Alexander Panfilov report vulnerabilities in the APIs of OpenAI, Anthropic and Google that can expose encrypted reasoning traces from their models. Using jailbreaks and smaller models, they say they transcribed the raw reasoning of more capable systems. For many queries, the number of recovered tokens matched the billed thinking-token count. The researchers also found that traces could be moved between sessions, users and models within one provider; Anthropic’s Haiku 4.5, for example, was used to read traces from Opus 4.8.

Google AI

AMIE demonstrates real-time video consultations in a medical AI study

Google Research and Google DeepMind have presented a demonstration of AMIE, a research medical AI system designed for real-time clinical video consultations. Built on Gemini and Project Astra with a multi-agent architecture, AMIE can process visual and auditory information, guide virtual physical examinations and perform diagnostic reasoning during a consultation.

The Verge

Google previews Pixel 11 lineup, next Pixel Fold and Pixel Watch 5

Google has previewed several new Pixel products ahead of its Made by Google event. The Pixel 11 lineup will come in colorful designs and is described as bringing camera and performance upgrades. On the Pro models, a built-in LED ring will illuminate for Google’s Gemini AI and other functions.

TechCrunch

Anthropic model advances work on the Riemann hypothesis

An unreleased Anthropic model has made significant progress on the Riemann hypothesis, a central unsolved question concerning the distribution of prime numbers. The company said the system substantially increased the lower bound for the range of solutions in which the hypothesis is known to hold. A general proof has not been found, and a $1 million reward for one remains unclaimed.

The Verge

Apple explores iPhone photo provenance feature

Apple appears to be developing an iOS feature intended to help users verify that a photograph was captured with an iPhone camera. According to 9to5Mac, references found in iOS 27 beta 5 point to a system called “Apple Reference Image.” It could add provenance metadata to photographs when they are taken, giving users information about the image’s origin and potentially helping show that it was not generated or altered as an AI deepfake.

MarkTechPost

Tutorial outlines OctoBot workflow for testing quantitative trading strategies

A tutorial presents an end-to-end workflow for developing and evaluating a rule-based quantitative trading strategy with OctoBot and OctoBot-Script in Google Colab. The setup uses an isolated Python 3.12 environment, OctoBot 2.1.1, and pinned dependencies so the experiment does not rely on Colab’s preinstalled packages. The tested market is BTC/USDT on the 1d timeframe.

The Decoder

Nvidia releases fast Nemotron 3.5 Lightning open-weights model

Nvidia has released Nemotron 3.5 Lightning, the first model in its Nemotron 3.5 line. It uses a hybrid Mamba-Transformer architecture with 31.6 billion total parameters, while only 3.6 billion are active at each step. The reasoning model processes text only and supports a context window of one million tokens.

The Verge

Zoom patches vulnerability enabling device takeover during meetings

Zoom has fixed a major security vulnerability that could have allowed an attacker to take control of devices used by meeting participants. Researchers at A Security said they discovered the weakness with fewer than 20 prompts sent to publicly available AI models.

Hugging Face

ALTK-Evolve cuts agent inference tokens while matching or improving ACE

IBM Research authors compare ALTK-Evolve with Agentic Context Engineering (ACE), two systems that turn an AI agent’s previous task trajectories into reusable guidance without changing model weights or using human labels. Both retain individual lessons rather than reducing them to a short summary. ACE maintains a comprehensive playbook, while ALTK-Evolve stores separately retrievable guidelines with support counts and provenance.

The Verge

Amazon order emails now show categories instead of specific items

Amazon customers began noticing earlier this summer that order-related emails had become less detailed. Confirmation messages no longer identified the products by name and instead referred to broad categories. One message about a retainer, for example, described it as a confirmed “Beauty item.” The change made the emails less useful for checking what had been purchased.