Daily AI News

OpenAI launches ChatGPT 5.5 Pro with breakthrough math capabilities

OpenAI unveiled ChatGPT 5.5 Pro, a new flagship model that delivered a PhD‑level breakthrough in number theory within an hour, according to Fields Medalist Timothy Gowers. The release showcases unprecedented reasoning and research assistance capabilities.

Read source →

NVIDIA releases Star Elastic: one checkpoint housing 30B, 23B, and 12B reasoning models

NVIDIA’s Star Elastic introduces a post‑training technique that nests three reasoning models (30 B, 23 B, 12 B) within a single checkpoint, enabling zero‑shot model slicing without separate training runs.

Read source →

9 Best AI Tools for Spec‑Driven Development in 2026

A roundup of the leading AI‑powered platforms—Kiro, BMAD, GSD, and others—that help engineers generate production‑ready code from structured specifications, streamlining the spec‑driven development workflow.

Read source →

OncoAgent: Dual‑Tier Multi‑Agent Framework for Privacy‑Preserving Oncology Decision Support

OncoAgent is a novel multi‑agent system that enables clinicians to receive AI‑driven oncology recommendations while preserving patient privacy through decentralized data processing.

Read source →

Google developers significantly misstate carbon emissions of proposed UK datacentres

Planning documents for two Google AI datacentres in the UK understated projected carbon emissions by a factor of five, raising concerns over climate reporting accuracy.

Read source →

Nvidia has already committed $40 B to equity AI deals this year

Nvidia disclosed that it has invested roughly $40 billion in equity stakes across AI startups and platforms in 2026, reinforcing its role as a dominant capital player in the AI ecosystem.

Read source →

Broadcom says it won’t build OpenAI’s custom chip unless Microsoft buys 40% of the supply

Broadcom is demanding a 40 % pre‑purchase commitment from Microsoft before proceeding with the custom AI chip designed for OpenAI, creating a potential financing bottleneck for the project.

Read source →

Google’s “Preferred Sources” feature risks amplifying low‑quality content in search

Google’s new “Preferred Sources” setting lets the company prioritize certain publishers in search results, a move critics argue could undermine the open web and favor proprietary AI‑generated content.

Read source →

Pseudoscientific emotion‑AI tools are infiltrating workplaces, says Atlantic report

A recent Atlantic investigation reveals that companies are deploying AI systems claiming to read employee emotions, despite lacking scientific validation, raising privacy and ethical concerns.

Read source →

Fields Medalist highlights ChatGPT 5.5 Pro’s unprecedented math research performance

Timothy Gowers, a Fields Medalist, reported that ChatGPT 5.5 Pro independently solved an open number‑theory problem, demonstrating AI’s emerging capability for original scientific contributions.

Read source →