Artificial intelligence

117 articles

Reporting on AI models, tools, research, infrastructure and policy.

Coverage: AI

Blackboard bold symbols defining the complex numbers in terms of the real numbers, written on a blackboard

NEWS AI Research

OpenAI Publishes a Batch of Lean-Checked Math Claims, and How Much Compute They Cost

OpenAI says it has published a broad range of new mathematical results from an internal frontier model, with computer-checkable Lean formalizations, per-result compute estimates of roughly three hours of ChatGPT Pro thinking, and release guidance shaped by consultation with an Institute for Advanced Study advisory group. The disclosure process is observable. The mathematics itself remains the company's own claim.

Diagram contrasting joint embedding, generative, and joint embedding predictive architectures, illustrating how content is mapped into a shared vector space

NEWS AI Infrastructure

Google's new open embedding model can search your voice memos and videos on-device, without uploading anything

Google DeepMind released EmbeddingGemma 2, a 740-million-parameter open model under Apache 2.0 that maps text, images, video and audio into one shared embedding space, enabling private offline media search on consumer hardware.

A typed contract form for participation in a television programme, dated 1975, shown as a printed legal agreement.

ANALYSIS AI Research

OpenAI trains its GPT-6 Astra agent on Ironclad's real contracting workflows, and the fine print matters

OpenAI announced a research collaboration with contract-software company Ironclad on October 6, 2026, using eleven real contracting workflows to train and evaluate its GPT-6 Astra agent. The reported gains are real findings from OpenAI's own evaluation, with simulated time estimates and no independent verification. Here is what the method is, what the numbers do and do not prove, and why the pattern may extend to other professions.

The JAMSTEC deep sea research ship Kairei at sea, the kind of research vessel whose shipboard measurements provide the sparse ocean observations behind sea surface pH reconstructions.

ANALYSIS AI for science

Why AI pH maps can look accurate while getting ocean carbon chemistry wrong: the REACT preprint, explained

A new arXiv preprint, REACT, argues that AI models reconstructing sea surface pH can minimize visible error while violating the carbon chemistry beneath. Its carbon-first framework reports smaller pH error and better chemical consistency, but only on simulation data.

Skyline of Abu Dhabi, United Arab Emirates, with the UAE flag visible

ANALYSIS Technology

A 7B model answers in Emirati dialect while giants answer in Modern Standard Arabic: what TII's Falcon-Emirati-7B shows about who AI understands locally

TII's new dialect-specialized model scored 0.52 on LLM-judged dialect fidelity in a vendor-run evaluation, against 0.05 or lower for competing models that mostly replied in Modern Standard Arabic. The result, if it replicates, says something important about how AI handles the languages people actually speak.

The entrance and main courtyard of Atlassian's Amsterdam office at Herengracht 124, photographed in 2009.

NEWS AI Applications

OpenAI and Atlassian expand partnership: GPT-6 family models will power agents across Rovo and the Teamwork Graph

Under a new agreement, OpenAI frontier models, including GPT-6 Astra and the GPT-5.6 series, will power agents across Atlassian's platform and Rovo, grounded in the Teamwork Graph enterprise context layer. The announcement pairs shipped integrations such as Codex plugins and the Atlassian ChatGPT extension with exploratory plans for assigning Jira work to AI agents and measuring impact through DX.

The front panel of an open-source Open Compute Project server, photographed in a data center

ANALYSIS AI Applications and Research

Auditing a black-box AI agent without its permission: the open-weight sidecar idea

A new arXiv preprint proposes scoring an agent's tool calls with a small open-weight model you control, recovering confidence signals that frontier APIs refuse to expose. The numbers are author-reported and unreplicated, but the access argument deserves attention.

Screenshot of the Undergarden, a fantasy underground dimension added to Minecraft by a community mod, showing custom biomes, vegetation and creatures integrated into the game world.

NEWS AI research

AI agents can now edit working game worlds, but most failed edits still look like success

A new preprint tests whether frontier AI coding agents can edit working games like Minecraft and Terraria without breaking them. The best configuration solves 78.2 percent of tasks, but most failed edits still build and load. The lesson for software teams: a passing build is not proof an AI edit did the right thing.

A text-to-image generated artwork of an astronaut riding a horse in the style of Hiroshige, produced with Stable Diffusion 3.5 Large.

ANALYSIS AI Safety Research

Keep It CALM: A Geometric Case That One Global 'Unsafe Direction' Cannot Fix Text-to-Image Safety Everywhere

A NeurIPS 2026 paper argues that the dominant design of training-free image safety filters, which removes one reusable unsafe subspace across all prompts, faces a structural coverage-selectivity trade-off. Its proposed fix, CALM, routes each prompt locally. This analysis explains the argument for general readers, and flags which parts are demonstrated versus self-reported.

The former London Stock Exchange building in the City of London, headquarters of the market where the 1,362 UK listed companies in the study trade.

ANALYSIS AI Policy

Almost half of UK annual reports now mention AI as a risk. Only 4.3 percent say anything substantive about it.

A new arXiv preprint classified 9,821 annual reports from 1,362 UK listed companies and found that AI risk mentions jumped from 2.8% to 41.2% in five years, while only 4.3% of 2025 reports contain disclosure the study deems substantive. The gap between boilerplate and substance is the story.

A rainbow trout swimming underwater, photographed at Gavins Point National Fish Hatchery in Yankton, South Dakota.

NEWS AI Research

AI reads trout swimming to decide when to feed, but the 96.67 percent result is unreviewed

A preprint from October 1, 2026 reports that a camera-derived activity measure and a fine-tuned language model can reach 96.67 percent decision accuracy for rainbow trout feeding, up from a 33.33 percent text-only baseline. The results are the authors' own, unreviewed, and limited to one species in research settings.

A NASA GEOS-5 simulation showing global atmospheric weather patterns over several days of simulated time

NEWS Research

MintFlow Preprint Reports Training-Free Way to Make Generative AI Respect Physics Without Distorting Its Output

A new arXiv preprint from researchers at UCLA and Brookhaven National Laboratory describes MintFlow, a training-free sampling method that nudges a generative AI model's trajectory as little as possible so results obey constraints like measurements and physical laws while staying close to the model's natural output. Results are the authors' reported claims in an unreviewed paper.

A NASA Hubble Space Telescope composite showing the Sombrero Galaxy at left with its faint stellar halo, and zoomed regions of halo stars at right.

EDUCATIONAL AI Applications, Research and Public Impact

The Sombrero Galaxy's Hidden Streams: A Seven-Day Exposure Shows a Galaxy Built from Shredded Smaller Ones

A seven-day amateur exposure from Namibia, featured as NASA's Astronomy Picture of the Day on October 5, 2026, reveals tidal streams and a faint stellar halo around the Sombrero Galaxy, M104. The observation shows that big galaxies carry the debris of the smaller galaxies they have devoured, and that this fossil record is readable from the ground.

European Union flags lined up in front of the Berlaymont building, seat of the European Commission, in Brussels.

NEWS AI Policy

OpenAI launches textGrain watermarking for EU text provenance rules, keeping its detector closed to the public

Starting today, API customers can opt in to text watermarking; EU ChatGPT and Codex output gets an invisible watermark in coming weeks; detector access goes first to approved researchers only. OpenAI's own numbers show why the public cannot check text itself yet.

Rows of server racks with blinking status lights inside a computer server room at NOIRLab headquarters in Tucson, Arizona, illustrating AI data center hardware capacity.

NEWS Blockchain Infrastructure & Verified Adoption

NVIDIA and Nscale ran 37% more GPUs on the same power budget: how dynamic power sharing turned stranded watts into AI capacity

A joint NVIDIA and Nscale evaluation reports that policy-governed power sharing let 192 GPUs run under a 264.4 kW budget sized for 140, with 49.2 percent more aggregate throughput and a measured cost in tail latency. All numbers are self-reported vendor measurements.

Die shot micrograph of an NVIDIA GM200 graphics processing unit, the chip used in the GeForce GTX 980 Ti

ANALYSIS AI Infrastructure

NVIDIA's OpenShell Puts the Locks Outside the AI Agent: How a Formal Policy Prover Changes Who Decides What an Agent May Touch

The open-source runtime enforces permissions outside the agent workload, keeps credentials out of reach, and uses formal logic to prove what a policy allows. Here is what is demonstrated, what is claimed, and what remains unproven.

Lines of code displayed on a laptop screen in a dark room, illustrating software security review

ANALYSIS AI Policy and Access

Arbitrum's New Security Program Puts AI Screening Ahead of Human Audits: What the Four Pillar Design Actually Does

The Arbitrum Foundation launched the Arbitrum Security Program on September 29, 2026, committing roughly 7.8 million dollars in USDC and ARB to a twelve-month program built on four pillars, the first of which is AI-assisted screening before a full human audit. This analysis explains how the machine layer fits into a DAO security pipeline, what stays human, and what the launch does not prove.

The Buckhorn Exchange, a historic Denver restaurant storefront, photographed in December 2025, illustrating the small-business Denver hospitality context of the story

NEWS AI Applications

A three-person Denver club says ChatGPT Work cut its liquor-license filing from four days to three hours: the numbers are OpenAI's own

OpenAI's October 1 customer story says a three-person Denver parent club saved 10-15 hours a week and cut grant and liquor-license paperwork from days to hours. Every number is vendor-published and self-reported. Here is what the page actually documents, what remains unverified, and what would prove it.