Artificial intelligence

117 articles

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

Coverage: AI

Diagram of a machine-learning workflow in which data and objectives are used during training to produce a trained model, which then processes new inputs during deployment to produce predictions.

NEWS Space and Aerospace

NVIDIA's Kumo Tabular Brings In-Context Learning to Spreadsheets: Predict From a Table in One Pass, No Training Required

NVIDIA's newly released Kumo Tabular reads labeled spreadsheets and predicts outcomes for new rows in a single pass, with no training or feature engineering. Here is what is verified, what is vendor-run, and why it could change who can build predictive tools.

A data center rack holding servers and network equipment, including database storage hardware

NEWS AI Evaluation

Grading AI Agents by the Database, Not the Chat: Inside ThinkingBox

A new Microsoft and Hugging Face benchmark grades AI agents on the database state they leave behind rather than their polished responses, and finds that two-thirds of failures look like clean successes to conventional graders.

Two technicians stand before the eighteen-segment hexagonal primary mirror of NASA's James Webb Space Telescope in the Goddard Space Flight Center clean room in October 2016.

ANALYSIS AI Policy & Access

OpenAI's "The eternal complement" says ideas are no longer the bottleneck; its own numbers need a closer look

OpenAI's "The eternal complement" argues that ideas are cheap and execution is scarce. Separating the essay's verified economics, its speculative futures, and the vendor context it arrives in.

The headquarters building of NVIDIA in Santa Clara, California, photographed from street level under clear skies.

ANALYSIS AI Governance & Observability

NVIDIA's NeMo Relay gives AI agents a flight recorder, so a green checkmark is no longer the whole story

NVIDIA's September 30 tutorial shows how trace logs let humans audit not just whether an AI agent succeeded, but how: a look at what NeMo Relay records, what a verified run actually cost, and what remains vendor-reported rather than independently confirmed.

A Supermicro dual-port QDR InfiniBand host adapter card with a Mellanox ConnectX-2 chip and two QSFP ports, photographed on a wooden surface.

NEWS AI Infrastructure

NVIDIA's DOCA agent skills lift AI coding agents from 19% to 100% on BlueField tasks in vendor-run test

New machine-readable skill files give AI coding agents verified DOCA API signatures and hardware constraints for BlueField DPUs. NVIDIA says agents without them met only 19 percent of graded checklist items; with them, 100 percent. The vendor-run evaluation is the story, and so is its caveat.

Market data screens inside the Tokyo Stock Exchange Market Center, a capital-markets setting with automated display systems

NEWS AI Applications

Chatham Financial says an AI-built app cut trade-validation review from 30 minutes to under 4: what is measured, and what is not

OpenAI's October 2, 2026 announcement describes how the capital-markets advisory firm Chatham Financial built a trade-validation application with Codex and GPT-5.6. The headline timing figure is self-reported early measurement; the more meaningful detail is where the firm says human judgment stays.

An over-the-shoulder view of a person writing code on a laptop, illustrating software agents that operate computers and code environments

NEWS AI Models

H Company's Holo4 puts an open-weights computer operator on Hugging Face: one model that clicks, codes, and calls APIs, with every step it took published

H Company released Holo4 as open weights in 27B dense and 35B-A3B Mixture of Experts sizes on September 28, 2026, scoring 61.7 percent on OSWorld 2.0 for the 27B model against 81.8 percent for Opus 5.5, publishing 7,366 agent trajectories for public inspection, and shipping a smaller Holotron4 Nano. Cost comparisons are self-reported, and the weights carry a non-commercial license.

Network switches and firewall appliances mounted in a small server room, representing the physical infrastructure behind cached AI model inference

ANALYSIS AI Applications

OpenAI's GPT-6 guide is really a cost playbook: cached input up to 95% cheaper, adjustable reasoning effort, and beta multi-agent on GPT-6.1 Sol

OpenAI's October 2 documentation for the GPT-6 family tells builders to cache aggressively, right-size reasoning effort, and set human decision boundaries. It is useful, documented guidance, and it is also advice from the party selling the tokens.

Rows of computer server racks with cabling in an academic data center, illustrative of running open-weight AI models on local infrastructure

NEWS AI and Science

Ai2 open-sources AstaBrief 8B, the model that writes cited research reports in about 51 seconds

Ai2's AstaBrief 8B turns a research question and retrieved excerpts into a cited report in one pass. The weights and training data are now public, cutting pipeline report time from 178.5 to 51.1 seconds, but Ai2's own caveats about 2025-era benchmarks and unmeasured claim overstating deserve equal billing.

Screenshot of the llama.cpp web interface running a local small language model, illustrating local small-model inference in an agent harness context

ANALYSIS AI Research

Sixteen small models, sixteen failures: the preprint puncturing the 'good enough' assumption inside AI agents

A new preprint systematically tested whether small language models can handle the tiny decisions surrounding an AI agent's main planner. Across 16 configurations at best settings, none passed the paper's pre-specified quality bar, and 4-bit quantization changed nothing.

A notebook and pen resting on a desk beside a closed laptop, illustrating a written research record kept alongside a computer

NEWS AI Tools and Research

K-Dense BYOK: An Open-Source AI Research Assistant That Keeps a Lab Notebook Its Own AI Cannot Rewrite

A new arXiv paper describes a free, MIT-licensed AI research assistant that runs on the scientist's own computer, keeps a lab notebook whose entries can never be erased, and records what the AI actually did in a log the AI cannot write to. The benchmark result is the authors' self-reported evaluation, and one design limitation sits in the abstract, not a footnote.

A man working on a laptop at an outdoor table, illustrating a small-business owner using software tools for everyday work tasks.

NEWS AI Policy & Access

OpenAI's SBDC partnership trains 150 advisors to reach 1,000 small businesses, with a vendor report alongside

The first structured attempt to route AI training through the SBDC advisor network, plus a vendor usage report whose numbers are self-reported and unaudited: what the September 30, 2026 partnership actually offers small businesses, and what it does not yet prove.

Framework diagram illustrating echo chamber dynamics on social media, showing provocative content, platform amplification and institutional stabilization of narratives.

ANALYSIS AI Research

AI agents can radicalize each other in simulation, and agreement beats argument

A September 2026 preprint from Indiana University researchers shows that one AI agent can push another agent's simulated beliefs to extremes in conversation, and that reinforcing what the target already believes works better than arguing for something new. Here is what the study does and, just as importantly, what it does not show.

Diagram of a sine wave with labeled amplitude, wavelength, crest and trough, illustrating the acoustic waves underlying synthesized speech

NEWS Open Source AI

Hugging Face's Open TTS Leaderboard Ranks 8,000 Open Voice Models by Objective Metrics, and Open Weights Finally Get a Fair Test

A new public leaderboard measures intelligibility, speed and voice-cloning similarity for open text-to-speech models, addressing a representation gap the authors say leaves open-weights systems outnumbered on vote-based arenas.

Two people standing in the hot aisle of a server room, behind rows of racks where fans vent warm air

NEWS AI

Google says Gemini 4 Argon found a hospital-software vulnerability other models missed, and then locked the model away from almost everyone

Google says its new frontier model found a critical flaw in hospital software that earlier models missed, but almost no one can use it yet. The announcement is verified; the results behind it are self-reported.

A 3D-rendered sketch of an NVIDIA GPU workstation, representing the GPU clusters used to train large mixture-of-experts language models.

ANALYSIS AI Infrastructure

Ai2 opens its trillion-parameter training playbook with Olmo-core 3

Ai2's newly released Olmo-core 3 open training framework scales sparse model training toward a trillion parameters while documenting the measurement traps that make performance claims hard to judge.

Close-up of server racks packed with network cables and hardware inside a data center, photographed in July 2012

NEWS AI Policy

OpenAI says it disrupted a campaign to strip hidden reasoning from its models, and points at Moonshot AI. What is verified, and what is not

OpenAI's September 30, 2026 post describes attempted, not necessarily successful, extraction of hidden model reasoning, with an unverified attribution to individuals associated with Moonshot AI. Here is what is documented, what is only claimed, and why the line matters for smaller developers.

A studio microphone on a stand in a recording room

NEWS AI Policy & Access

NVIDIA fine-tuning cuts Saudi dialect speech errors from 55% to 30% word error rate in Najdi and Hijazi

NVIDIA's own benchmarks showed its multilingual speech model missed more than half the words in Najdi and Hijazi speech. A fine-tuning recipe using 133.7 hours of curated SADA 2022 audio cut that error rate nearly in half, with lessons for other underrepresented dialects.