Full-text search across 70304 articles. Combine with topic and date filters; results sorted by relevance.
GPT-6 Astra High won a 10-model round-robin Pokémon Showdown tournament, taking 75 of 90 games, while Qwen 3.8 27B finished last, according to an experiment run by the cofounder of an inference company who spent $1,000 o…
The Allen Institute for AI (Ai2) released ToyBlocks on November 6th, an open simulation framework in which a robot arm pushes, flips and stacks blocks on a tabletop, alongside code, data and two models for repeatable tes…
Cerbère AG published a November 6, 2026 post arguing that cloud-based AI agent guardrails are self-defeating because they require customers to send prompts, tool calls, secrets and API responses to a third party, and pos…
A developer outlined two approaches for pulling data from Model Context Protocol (MCP) servers into a frontend web app: running the official TypeScript SDK directly in the browser over Streamable HTTP, or routing calls t…
Developer ehildt released tripleF, an open-source agentic AI chat workbench that runs free open models fully locally via Ollama or through Ollama Cloud, with no usage meter or proprietary lock-in. The project, described …
A new Behavior-Conditioned Symbolic Synthesis (BCSS) framework, detailed in an arXiv paper (2611.07088), automatically extracts deterministic, interpretable symbolic programs that verifiably reproduce the reasoning pathw…
A developer running a multi-agent team for code, docs, PR review and deployment reports that agents improve through shared, structured experience rather than individual prompting. Using Octo, an open-source human-AI coll…
A project called Repairer, published by gno.security, has released an open-source Go program that runs on each server and applies a signed stopgap the day a Linux exploit goes public, then automatically removes it once t…
A new arXiv paper (2611.07055v1) introduces competitive flow matching, a training method that adds a randomly drawn low-dimensional conditioning code z and samples K values of z per example so K candidate velocities comp…
A new arXiv paper (2611.07063v1) introduces SD-LoRA, a structural distillation framework that trains a heterogeneous graph teacher over high-resolution ViT tokens and transfers its structural bias into a lightweight LoRA…
Researchers introduced TB-MathModel, a terminal-style benchmark that makes 18 real CUMCM and MCM contest problems from 2023–2025 agentic and verifiable, scoring agents' single JSON artifacts against reproductions of Outs…
Researchers posting arXiv paper 2611.07081v1 introduced Wide-View Gaussian Splatting (WV-GS), a feed-forward Gaussian-splatting framework that synthesizes a geometrically consistent wide view from multiple perspective ca…
Anthropic's privacy hook in Claude Code v2.1.317, enabled by default, fails to block 8 of 23 email-leak shapes tested by Frank Chu, with a reproduction on v2.1.318 filed as GitHub issue 107757. The hook, cc-plugin-privac…
Cloudflare has added a documented 503 response to the POST /accounts/{account_id}/ai-gateway/billing/topup endpoint and its /config variant, according to the company's published API specification. The change means code t…
A developer published a JavaScript module, timeline.mjs, that maps between source and output timestamps for a sequence of hard cuts, handling reordered and repeated source ranges and per-range playback speed. The mapper'…
Stroq has introduced a zero-telemetry "causal firewall" that traces terminal executions by autonomous coding agents back to the untrusted text that provoked them, aiming to block indirect prompt injections before rogue t…
Mistral released the weights for Mistral Large 4, a 1.05 trillion-parameter mixture-of-experts model nicknamed "Le Chonk," on October 29, making it the largest open-weight model published by any European or American lab.…
Kapwing launched a Model Context Protocol (MCP) server that lets Claude Web, Claude Desktop, Claude Code, Cowork, and ChatGPT users upload footage, pull exports, generate and edit media, and subtitle, dub, and translate …
A developer built One-Step SEO, an open-source, zero-dependency Node.js CLI that audits a website's traditional SEO and Generative Engine Optimization (GEO) readiness and reports the two as separate scores. Run via npx, …
Baidu released ERNIE-5.5-Flash-Base, an open foundation model with 280 billion total parameters and 14 billion active per input, on Hugging Face in FP8 and BF16, deliberately shipped without fine-tuning or post-training.…