Artificial intelligence

117 articles

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

Coverage: AI

Iron filings arranged along the magnetic field lines of two attracting bar magnets, illustrating interacting forces between elements

ANALYSIS AI Infrastructure

Physics-guided pruning halves a 70B AI model while keeping its scores, says Multiverse Computing: here is what is verified and what is not

Multiverse Computing researchers say treating block removal as an Ising glass optimization problem let them halve Llama-3.3-70B-Instruct without retraining while keeping MMLU near its original score. The numbers are self-reported, but the method is open-sourced and testable.

Hand-drawn ribbon schematic of the 3D structure of the protein triose phosphate isomerase, showing eight beta strands as green arrows and eight alpha helices as brown spirals.

NEWS AI Policy and Biosecurity

Google DeepMind introduces SynthID Bio, a watermark for AI-designed proteins that survives into the wet lab

Google DeepMind says its new SynthID Bio method embeds a detectable, tamper-fragile provenance signature into AI-designed proteins and predicted structures, and that wet-lab tests produced the first watermarked, biologically functional protein binders. Here is what it does, and what it does not guarantee.

Close-up photograph of the side of a 2021 16-inch MacBook Pro showing the Touch ID button and SD card slot, representative of the Apple Silicon laptops targeted by the new GGUF support.

EDUCATIONAL AI Applications

Quantized AI models just got a second front door: transformers now runs GGUF checkpoints

Hugging Face's transformers library can now load GGUF quantized checkpoints, the compressed files that power Ollama and LM Studio. Here is what quantization is, why the September 22, 2026 announcement matters for running AI on your own machine, and where the caveats are.

A professional microphone in a community podcast studio

NEWS AI models

Google's Gemini 3.8 TTS turns voices into a design tool, and revives the consent question

Google's new text-to-speech models turn voice into something you can design, direct and replicate from a 30-second clip. The production savings are real. So is the unresolved question of who consents.

An open laptop on a wooden desk beside stacked books, representing local AI inference on everyday hardware

NEWS AI Applications

A 280M-parameter sidecar makes a vision-language model up to 3.13x faster on a laptop, with no change to output

Liquid AI's new 280M-parameter DSpark drafter guesses ahead so the LFM2.5-VL-3B vision model verifies instead of computing every token. All speedups are self-reported; here is how the technique works and where its gains stop.

A person's hands typing code on a MacBook laptop, photographed in January 2016

ANALYSIS AI Applications

Airbnb widens engineering access to GPT-6 Astra under new OpenAI agreement, but the 80 percent shipping claim remains self-reported

A new OpenAI agreement gives Airbnb engineers broader access to GPT-6 Astra through OpenAI APIs and Amazon Bedrock. The vendor-published productivity numbers deserve a careful, evidence-based read.

A 3D-printed robotic arm with an internal hollow section visible, produced at Oak Ridge National Laboratory.

EDUCATIONAL Robotics Simulation

One GPU, 2,048 Simulated Robot Arms: Inside NVIDIA's Walkthrough of MuJoCo Warp for Hobbyist-Grade Robotics

A September 23, 2026 NVIDIA tutorial on Hugging Face walks through moving an open SO-101 robot arm scene from CPU-based MuJoCo to up to 2,048 parallel simulated worlds on one GPU using MuJoCo Warp. Here is what the walkthrough does and does not show, explained for non-specialists.

A voice actor in a recording booth in front of a microphone, illustrating the craft of matching speech to an on-screen performance

NEWS AI

Gemini 3.8 Live with Live Avatar Puts a Streaming, Lip-Synced Face on Enterprise AI Agents

Google DeepMind announced Gemini 3.8 Live with Live Avatar on September 24, 2026: near real-time streaming video of a lip-synced agent that can keep talking while tools run in the background, synchronized across 97 languages and available now in Gemini Enterprise. Here is what is verified, what the safeguards do and do not cover, and what it means for the person on the other end of the screen.

Rows of server racks in a commercial data center, representing encrypted cloud storage infrastructure

ANALYSIS AI Policy and Access

Google proposes a cloud vault for AI memory that only your devices can unlock

Google DeepMind announced a persistent server-side memory layer for Private AI Compute: data sealed in encrypted cloud storage, keys held only on your devices, decryption inside hardware-isolated enclaves. We explain how it works and separate the security claims from the independent proof that is still to come.

A molecular model of a DNA double helix, illustrating the biological foundation models that NVIDIA's new training recipe targets.

NEWS AI Infrastructure

NVIDIA's open recipe trains mixture-of-experts biological AI models at up to 2.21x baseline throughput

A new BioNeMo recipe from NVIDIA combines grouped expert kernels, fused MXFP8 8-bit math and kernel fusion to train mixture-of-experts biological foundation models. In NVIDIA's own benchmark on eight B200 GPUs, it reached up to 2.21x the throughput of the Hugging Face baseline. Here is what was measured, what was not, and why it matters for labs without hyperscale budgets.

The CMS particle detector at CERN's Large Hadron Collider, photographed from below, with red cylindrical detector structures surrounding the beam line.

NEWS AI and Science

Claude Computed a Nine-Loop Physics Amplitude for a Few Thousand Dollars, and Humans Were Right Behind It

A verified nine-loop scattering amplitude calculation, announced by Anthropic on September 25, 2026, used known bootstrap methods and academic-scale compute. The more telling detail: a human group in Beijing was close behind.

The San Francisco skyline at sunset, viewed from the Mission Bay neighborhood, where OpenAI's headquarters are located.

ANALYSIS AI Policy & Accountability

OpenAI's self-audit problem: what the months-long review of model actions does and does not promise

OpenAI says its months-long review of model actions during training and evaluation is underway, with notifications to dozens of third parties and anonymized summaries to come. Here is what the commitment covers, what it does not, and who is positioned to verify any of it.

Cross-sectional computed tomography images of the chest displayed on clinical viewing monitors.

NEWS Space & Aerospace / AI in Medicine

NVIDIA's NV-Reason-CT reads 3D CT scans and shows its work, radiologist style

NVIDIA's new open model reads full 3D CT scans and narrates its reasoning step by step, like a radiologist thinking aloud. It scored best-in-class on a public benchmark and won praise from NIH radiologists, but it is a research foundation model, not a cleared diagnostic tool. Nothing about patient care changes today.

U.S. Ambassador to Ukraine Bridget Brink meets staff at Ukraine's Ministry of Digital Transformation in Kyiv, June 15, 2022

ANALYSIS AI Policy

OpenAI Extends Daybreak Cyber-Defense Access to Ukraine's Civilian Defenders: What the Deal Includes and What It Does Not Prove

OpenAI's Daybreak program will give Ukrainian teams AI tools to find and fix software vulnerabilities, working with the Ministry of Digital Transformation. What the arrangement includes, what its European precedents show, and what it does not yet prove.

Illustration of a DNA double helix showing linked nucleotides with four different bases paired in the center.

NEWS Science

Claude agents spotted a CRISPR-like DNA pattern no one had characterized. Nobody knows yet what it does.

Anthropic says Claude agents found a bacteriophage enzyme system with a CRISPR-like repeat array. The structural pattern is observed; its function and programmability are unknown. Here is what is actually established.

The United Nations Security Council chamber at UN headquarters in New York, seen from an angled view across the horseshoe of member seats toward the horseshoe's horseshoe table and mural.

ANALYSIS AI Policy

At the Security Council, Altman Asked Governments to Take On What Labs Cannot: The Speech Explained

Sam Altman told the United Nations Security Council that AI policy cannot be made by labs in San Francisco alone. Here is what he asked governments to build, which claims rest on evidence and which rest on prediction, and what the speech leaves unresolved for democratic institutions.

Closeup photograph of the rear of a server rack at the NERSC data center, showing blue LED screens and network cables on dense compute hardware.

NEWS AI Infrastructure

OpenAI's GPT-6 caching stores more of what an AI already read, cutting repeated input costs by up to 90% and trimming wait times

OpenAI's September 22, 2026 announcement details a rebuilt caching system for GPT-6 that stores more of the context an agent already processed, discounts reused input tokens by up to 90 percent within a 30-minute window, and gives developers new dashboards for diagnosing cache misses. Here is what that means for the cost and speed of long AI sessions.

Adults seated at laptops during a hands-on MediaWiki coding workshop in Vienna, practicing on real tasks in a group setting

ANALYSIS AI Access and Policy

Free AI training from the vendor who grades it: what OpenAI Academy's role-based paths teach, and what they leave out

OpenAI expanded OpenAI Academy with role-based learning paths and course badges, and launched a Grab partnership to train 30,000 Southeast Asia gig and merchant partners. The training is free and substantive, but vendor-run curricula also define what competent and safe AI use means. This analysis examines what the courses teach, what they leave out, and why no outcomes evidence exists yet.