The future leaves
signals in time.

Hugging Face lists FairStress, described as a deeper evaluation of fairness in LLM decision-making. No further context was provided, so the specific methods and findings are not stated.

The adoption of artificial intelligence in the tech sector, driven largely by Silicon Valley, shows trends that other industries can learn from. As generative AI and automation tools mature, the shifts currently affecting software engineering and product management are expected to spread.

OpenAI says Terminal X's Eval Agent grades outputs against a 54-criterion rubric and, using GPT-6.1 Sol, matches senior investor judgment 88.6% of the time while catching 25% more deviances from a fund's mandate.

OpenAI is improving ChatGPT's capabilities across low-resource programming languages. This improvement includes new evaluations for tasks such as software engineering, SQLite, and COBOL migration.

Brazilian bank Itaú is applying its AI First strategy to customers and employees through ia.i, a conversational interface, and the IARA governance platform, operating in one of Brazil's most regulated sectors.

An NVIDIA Technical Blog post describes using RTX Spark to code, test, and deploy on a single machine. It notes that AI agents are changing how software is built, with developers shifting from writing every line of code by hand to directing agents that write most of the code; the provided context is truncated.

According to OpenAI, Localiza uses ChatGPT Enterprise, the OpenAI API, Codex, and more than 60 specialized agents to reorganize work and build AI-native mobility experiences.

OpenAI introduced the Rosalind Academic Plan, which offers eligible academic labs sponsored access to GPT-Rosalind and tools to advance life sciences research.

Meta has launched Muse Spark Open, a coding-optimized 816B mixture-of-experts (MoE) model. This model features 42B active parameters and a 1M token context window.

In a technical blog post, NVIDIA describes Nemotron 3.5 Super VL as powering specialized agents with multimodal reasoning. The post notes that always-on agents are taking on longer, more complex enterprise workflows, where a single request can involve reading a document and checking information across sources.

Muse Spark Open from Meta is presented as an open-weight frontier multimodal mixture-of-experts model, per the headline. The feed provides no further details beyond this description.

An NVIDIA Technical Blog post describes using NVIDIA NeMo Gym to optimize agent reliability and tool-calling efficiency for chip design. It notes AI agents can deliver notable productivity gains in complex chip design workflows but tend to produce plausible yet incorrect artifacts; full method details are not included in the excerpt.

OpenAI's feed describes how Amp uses OpenAI models — GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT — to push AI agents' capabilities beyond coding tasks. No further outcome details are provided in the item.

Anthropic's large language model Claude reportedly identified a curious CRISPR-like stretch of DNA in less than a day, but even Anthropic's scientists are unsure what the pattern does.

According to OpenAI, AI lab Third Moment Research runs 10,000 agents for more than 36 hours using GPT-6 Astra, GPT-6.1 Sol, Codex, and Daybreak for cybersecurity; the lab says its agents can focus for days and improve with use.

A new analysis from OpenAI shows that adoption of agentic AI is growing among EU enterprises and small businesses, and outlines where deeper use could unlock more value.

WIRED reports that chatbot prompts allegedly show an OpenAI engineer's efforts to set up a covert listening device days before one was found in a meeting room used by senior executives. The report describes allegations, not confirmed findings.