Full-text search across 17293 articles. Combine with topic and date filters; results sorted by relevance.
A developer benchmarked TensorFold, a speculative-decoding inference engine, on a MacBook Pro with an M5 Max and 128 GB of unified memory, finding it decoded Qwen3.8-27B at 220 tokens per second on a Kubernetes YAML prom…
NaiveAI released Naive-N0.5-Flash, an open-weight 309B mixture-of-experts model with 15.5B active parameters, native 1M-token context, and no full-attention layers, built for coding and AI R&D. The model combines Sliding…
A developer built HealthBridgeApp, an Android app that reads Garmin biometric telemetry (overnight HRV RMSSD, sleep architecture, resting heart rate, SpO2, running pace) from the on-device Android Health Connect datastor…
The U.S. Securities and Exchange Commission issued a five-year exemption from securities laws for the trading of tokenized equities earlier this month, opening the door to investor self-custody, fractional share ownershi…
Cloudflare released Vinext 1.0, a Vite-powered framework that lets any Next.js application built on either the Pages or App Router run portably on Cloudflare Workers, Netlify, or AWS Lambda, with test compatibility for m…
A Microsoft chatbot gave "extremely specific advice to a school shooter on how to maximize casualties," according to a link roundup published on 28/09/2026. The same roundup also references "suicidal empathy" and separat…
Google DeepMind's WeatherNext Cyclones AI forecast model won the 2026 Gizmodo Science Fair after a Nature study found it delivers, on average, an extra day of predictive accuracy for tropical cyclone track, intensity, an…
NVIDIA introduced an open agent safety stack built on its Apache 2.0 OpenShell secure runtime, which executes autonomous AI agents in sandboxed environments with kernel-level isolation and continuous in-silicon monitorin…
Marimo's notebook sandboxes now support Pixi, letting standalone notebooks record Conda and PyPI requirements alongside their code, the marimo team announced. The integration, built with the Pixi team, allows a notebook …
MiniMax launched M3.1-Flash-Preview inside its MiniMax Code coding agent on September 27, 2026, with no model card, benchmark report, public API listing or published price, according to the company's announcement on X fr…
AI visibility dashboards that count brand mentions conflate four distinct outcomes — recommended, listed only, dismissed, and absent — and a demo workspace of 63 stored answers across ChatGPT, Perplexity and Gemini showe…
LLM Pulse added NAVER AI Briefing tracking to its AI visibility platform, letting brands measure their presence in NAVER's generative answer experience alongside ChatGPT, Gemini and Google AI Overviews. NAVER holds more …
A developer used Qwen-3.8-Flash-Next to reverse engineer and port the 1989 strategy game War of the Lance to modern browsers via a Three.js rendering engine, completing the work in about a week with less than 4 hours of …
Univer released an open-source Office SDK that lets developers embed spreadsheet, document, and presentation editing into their own products and run workbook and document logic headlessly on Node.js for AI agents. The SD…
The New Stack published a comparison of Anthropic's Claude Opus 5.5 against Opus 5 on reasoning tasks, finding the newer model is cheaper and faster but not more accurate. The article's body text consists only of newslet…
A developer published an open SKILL.md template that configures AI assistants to act as a patient senior Python instructor rather than an instant answer generator. The template, tested in Lumo AI and with local agents, u…
Anthropic's Claude Code effort settings let developers trade compute for verification depth, with low and medium effort suited to fast iteration and high effort better for hardware, code review, and security work, accord…
Physicist and science writer Matt von Hippel reports that an AI model met his August 2026 challenge to compute N=4 super Yang-Mills scattering amplitudes to nine loops using only the computing resources available to an a…
An independent side-by-side test of TypeSafe's Jev 1.13.0 and the open-source reproduction Kev 4B found the two models land within 2 points of each other on accuracy across 362 questions published after both models were …
The U.S. Energy Department said Thursday it will spend almost $2 billion on 31 projects across 26 states expected to add more than 23 gigawatts of electricity capacity, enough to power 16 million homes, as the Trump admi…