Full-text search across 70349 articles. Combine with topic and date filters; results sorted by relevance.
Researchers introduced EgoHand, a monocular 3D hand reconstruction model that achieves state-of-the-art performance on egocentric video, trained on a curated corpus of 14 million images combining synthetic data, in-the-w…
A new arXiv paper (2610.30355v1) introduces CiteAnchor, a source-aware grounding framework for detecting hallucinated citations, and CiteResolve, a benchmark of 423 real-world hallucinated citations annotated for source …
Researchers released Fig (Finding Interactions in Games), an agentic system that performs targeted elicitation interventions on game environments to test multiagent behavior, with code available at github.com/fleet-ai/fi…
DiffResearch, an open multi-agent deep research framework released under the Apache-2.0 license, runs a diffusion language model at every stage and is roughly 6 times faster end-to-end than a comparable autoregressive ba…
A new arXiv paper (2610.28945v1) proposes a test-time framework, inspired by Model Predictive Path Integral (MPPI) control, that lets large language model agents propose multiple candidate actions at each step and imagin…
Microsoft has open-sourced Scope, a self-hosted platform that runs AI coding agents against a developer's SDK, MCP server, or CLI and shows exactly where those agents fail. The tool is aimed at letting developers test wh…
Vercel added a `vercel feedback` command to its Vercel CLI, letting developers send bugs, friction reports and feature requests to the Vercel team directly from the terminal. The command accepts `-f` or `--field` flags w…
An unofficial, reverse-engineered MCP server for Onshape CAD exposes the platform's native editor commands and browser-authenticated REST calls to AI clients such as Codex and Claude Code through a local stdio API, requi…
Cloudflare cut its AI code reviewer's cost per review from $2.43 to $0.77 and median review time from 4m 09s to 1m 32s, a 68% reduction in inference bill, by using shared prompt-cache prefixes, fewer LLM calls per sessio…
A developer documented a debugging approach for multi-tenant RAG chatbots that answer from the wrong merchant's documents, tracing the root cause to a missing or misspelled tenant filter in the retrieval stage rather tha…
Amazon's MadPot honeypot network recorded the earliest PHP2Shell exploit attempt on Oct. 17, four days before PHP's maintainers shipped fixes for CVE-2026-109618 and CVE-2026-109619 on Wednesday, CJ Moses, chief informat…
A critical 9.2 CVSS remote code execution flaw dubbed PHP2Shell, tracked as CVE-2026-109618 and CVE-2026-109619, affects PHP versions from 5.4 through 8.5 and allows unauthenticated attackers to execute arbitrary shell c…
A German court ruled in favor of a man who sued Google for invasion of privacy after Google Maps and Google Earth published aerial images of his property without his consent, according to DW. The roundup also notes Googl…
A developer built InternLog, a small app for tracking internship work, weekly reports, and progress toward hour goals, using AI assistance within Replit. The developer reported that AI accelerated ideation and feature wo…
A developer who has spent years rolling out governed conversational analytics on Looker and its semantic layer, most recently exposing it to AI agents via the Model Context Protocol (MCP), argues that the gap between LLM…
Pactum, led by co-founder and CEO Kaspar Korjus, launched its Intake Agent on October 21, an AI agent that converts employees' plain-language purchase requests into coded requisitions and routes them into sourcing and ne…
Security researcher Sammy Hajhamid, who goes by pepsipu online, released ns2jb, a software-only jailbreak for the Nintendo Switch 2 that runs on the latest 23.0.1 firmware, using OpenAI's Codex to do much of the work. Th…
Wallaby Token has released reconcile.py, a zero-dependency audit script in its open-source wallaby-agent-rules repository that diffs an agent's dated project log against its current-state file to flag contradictions and …
A new benchmark called Quant-Arena evaluates large language model agents on the full iterative loop of quantitative trading research — proposing a market hypothesis, implementing it as an executable signal, testing it on…
Google Cloud launched self-learning context enrichment and a reimagined Knowledge Studio, both in preview, in Knowledge Catalog, its context engine for agents. The Gemini-powered enrichment analyzes agent and user intera…