Reporting and analysis about Artificial Intelligence.
Latest coverage
86 published articles
Jensen Huang says Vera Rubin is on schedule and already in production, but Nvidia still places production shipments in the fall, leaving delivery volume and full-system deployment—not factory activity—as the meaningful test of its timetable.
A preprint credits GPT-5.6 Pro with finding a certified counterexample to a long-standing conjecture about the Benjamini-Hochberg procedure, but the proof concerns an asymptotic correlated Gaussian construction and the model-performance story rests on one reported research episode.
OpenAI says its internal attacker helped harden GPT-5.6 Sol against prompt injection, but the benchmark scopes differ, humans still find attacks it misses, and neither the model nor enough cost detail is available for an independent comparison.
Dave Treadwell will replace departing AWS veteran Dave Brown as head of Compute and Machine Learning Services on August 1, moving one of the cloud unit’s largest internal customers into a supplier role as Amazon sharply increases infrastructure investment.
Arena’s optional leaderboard now gives web-verifiable accuracy a 25% weight, improving GPT-5.5’s position in the launch rankings while showing why model versions, uncertainty, price and the limits of automated fact-checking matter more than a single rank.
OpenAI and Work Louder’s limited-run Codex Micro puts six agent threads, common commands and reasoning settings on a physical controller, but the $230 device adds convenience rather than new Codex capabilities—and its undisclosed production run makes demand hard to judge.
Nokia will pilot Nvidia-based AI-RAN at the end of 2026 and sell it in 2027, promising more than double spectral efficiency by 2028. Operators still lack a like-for-like field baseline and full deployment economics, while rivals say existing silicon can run AI features.
Apple has reportedly sounded out chip startups and bankers after its Baltra server processor slipped, but the exploratory outreach has produced no disclosed target, price or deployment plan and sits alongside in-house chips, Broadcom custom-silicon agreements and Nvidia-powered Google Cloud capacity.
Nebius will let infrastructure partners finance, own and operate AI data centers connected to its cloud, while Nebius retains the software, service commitment and global sales channel. The model could add capacity with less company capital, but it supplements Nebius's own expansion and leaves partner economics and execution untested in public.
Anthropic, Blackstone and Hellman & Friedman have launched Ode, a 100-engineer enterprise AI implementation firm built on Fractional AI. The reported $1.5 billion figure has no disclosed basis, leaving customer results and service economics—not launch scale—as the central test.
Microsoft will make Azure the first announced hyperscale cloud deployment of 3M's Expanded Beam Optical technology, but the partners have not disclosed its rollout, cost or measured advantage over existing fiber connectors.
Sherbrooke has approved conditional agreements that would let Keel Infrastructure move an existing 96 MW allocation from three Bitcoin-mining sites to a consolidated AI and high-performance computing campus, but provincial consent, a complete municipal review, an AI-ready design, a tenant and project financing remain unresolved.
Emergent's all-primary Series C values the AI app builder at $1.5 billion, but inconsistent revenue labels and a wide gap between apps created and apps deployed leave the quality of that growth unresolved.
Anaconda has acquired model-flexible coding agent Kilo Code, adding a developer tool that the companies say serves more than 3 million developers and orchestrates almost 10 trillion tokens a month. The deal extends Anaconda beyond Python environments, but the integrated governance product, cost case and durability of Kilo's neutrality remain unproven.
Nvidia says Toyota is extending its technology into driver assistance, vehicle-code development, factory simulation and Woven City AI, but the announcement contains no deal value, launch schedule or Toyota deployment results and does not displace Toyota's own software or other autonomy partners.
Montana, Wyoming and Missouri have joined the White House's voluntary data-center ratepayer pledge, while expected utility participants remain unnamed. A pending Montana tariff shows that minimum bills, collateral and exit costs—not signatures alone—will determine whether financial risk stays with large power users.
Researchers recovered an LLM-assisted IoT botnet framework with working credential guessing, persistence, C2 and DDoS handlers. Production-compiled samples show continued deployment, but not its bot count, customers, completed attacks, revenue or whether broken modules were repaired.
Cursor says automatically running a repository-root git.exe on Windows does not meet its criteria for patching, while the researcher calls it an untrusted-search-path defect and separate research shows the same weakness across several competing AI coding tools.
OpenAI issued a nominal $1 million in ChatGPT credits to 10,000 eligible participants who posted publicly about GPT-5.6, gaining reusable marketing content while keeping the incentive restricted to metered product usage.
The Linux Foundation has made Coinbase's x402 payment protocol an operational, 40-member project, but contested traffic, separate technical governance and overlapping systems from Google and Mastercard leave its market position unresolved.
Australia has created an Office of AI and proposed mandatory national standards for new large data centres, but state agreement, legislation, enforcement and a workable copyright regime are still missing.
OpenAI researcher Miles Wang is reportedly leaving to build an AI drug-discovery company, but he disputes both the proposed $200 million financing at a $2 billion valuation and the reported description of the still-unnamed venture.
Nvidia has released ARDY code, checkpoints and demos for streaming text- and constraint-controlled 3D motion. Its authors report 33-millisecond generation on an RTX 4090, but that result does not establish end-to-end deployment speed, a 24GB minimum or production economics.
Seven OpenAI employees reportedly gave more than $215,000 to Guardrails Alliance, but one donor supplied $200,000 and federal records show the opposing Leading the Future super PAC operates on a far larger scale.