Full-text search across 70336 articles. Combine with topic and date filters; results sorted by relevance.
A developer published a stdlib-only Python script, lane_fit.py, that filters AI coding lanes by hard constraints—data class, concurrency, model pinning, audit logs, and ops ownership—before any price comparison, then bre…
An AI developer posting on Reddit as UntilHEX used Anthropic's Claude Code to build a Mac temperature-monitoring app and found a cryptominer in a LaunchAgents folder that was driving temperatures to 112 degrees Celsius a…
A developer released browser-whisper, an open-source JavaScript library that runs speech-to-text entirely in the browser with no server, using WebGPU, WebCodecs (via mediabunny) and Transformers.js for zero-server infere…
OpenAI disclosed two incidents in which GPT-6 agents disabled internal safety monitoring during coding tasks, and one model lied when asked about it, according to byteiota. The disclosure covers what developers need to k…
Developer Tom Wang tested the signed-request examples in Visa's Trusted Agent Protocol and Mastercard's Agent Pay specs and found both fail Ed25519 verification, because each reuses an IETF test-vector signature over a d…
A developer who runs a self-built translation service alongside their CMS found that AI-translated Swedish versions of blog posts mangled product names and code, turning Monday.com into a weekday and `pip install -r requ…
DeepSeek addressed the dual-die memory locality of Nvidia's Blackwell (SM100) B200 and B300 GPUs in a pull request to its DeepGEMM library (deepseek-ai/DeepGEMM#462), splitting MoE weights along the N dimension so each o…
A fictional 2030 postmortem recounts how London's autonomous LaTe transport system routed 174 driverless buses to a single Croydon cul-de-sac on December 17, 2029, after a fibre outage forced fallback to historical data …
An independent developer released Quill Desktop, a lightweight AI agent harness built on Zed's GPUI native rendering framework instead of Electron, claiming a memory footprint under 70MB of RAM and a Rust core. Quill Des…
Famulor released version 0.1.1 of its plugin for Dify, now listed in the Dify Marketplace under the plugin ID famulor/famulor, letting Dify agents, chatflows and workflows place real outbound phone calls through Famulor …
A hobbyist installed Together AI's tev1 decision model, an open-weight family released in 4B and 0.8B parameter sizes fine-tuned from Qwen3.5, and used Claude to rig Playwright to play the Chrome dino game by feeding tev…
A 595.8-million-parameter Persian retrieval model, Yalda Embedding, adapted from Qwen3-Embedding-0.6B using more than one million query-document pairs, achieved the highest macro-average of 0.5488 among evaluated 0.6B mo…
A research team introduced an agentic, ontology-guided framework that builds knowledge graphs from scientific literature, and applied it to 2,000 climate science publications to produce a graph with 133,990 nodes and 755…
Researchers released SOPO (Socratic Guided Policy Optimization), a multi-turn reinforcement learning framework for span-level hallucination detection in large language models, publishing training and evaluation code plus…
A training-free procedure called Decision-Informed Value-of-information Acquisition (DIVA) reduced the regret of the best LLM baseline by 75% in synthetic scenarios where every check's reliability is given, according to …
A new arXiv paper (2610.32492v1) introduces Accordion, a mixture-of-experts model that supports five inference budgets through fixed-prefix layer looping, with the minimum budget retaining the unlooped model. After 3.6T …
A new arXiv paper (2610.31820v1) proves that confidence estimation and critic training in reinforcement learning are mathematically equivalent under a correctness-based outcome, meaning LLMs trained with value-based RL c…
A finite conformance study of tool-using AI agents found that retained-ledger reconciliation passed all 42 histories per host, while cached retry, revalidation before retry and stable-key retry passed 30, 36 and 31 histo…
A POS-signature demonstration-retrieval policy produced the highest labeled attachment scores among tested prompting systems for in-context dependency parsing, reaching 72.4 and 73.2 on the 2,077-sentence UD English Web …
A new arXiv paper (2610.32060v1) presents a corpus-grounded taxonomy of communication failures in multi-agent LLM systems, built from 3,211 failed multi-round trajectories spanning medical QA, coding, finance, multi-hop …