Full-text search across 26459 articles. Combine with topic and date filters; results sorted by relevance.
A developer built VCIS-Bench (Visual Constraint & Instruction Simulation), a synthetic 7×7 grid puzzle benchmark testing whether vision-language models can perform step-by-step spatial verification, and evaluated Claude …
OPSWAT released MetaDefender Aether v3.3.0, adding LLM Scanning that analyzes pretrained model files for hidden threats across Layer 2 static analysis and Layer 3 dynamic analysis. Static analysis parses 7 LLM file forma…
CrowdStrike Intelligence disclosed that a financially motivated, suspected Chinese-speaking threat actor used ARTEX, an open-source agentic penetration testing tool developed in China by Autumn-27, to breach South Korean…
Developer rbinar released cli-dispatch, an MIT-licensed Claude Code plugin that delegates coding tasks from a Claude session to five external CLI workers: DeepSeek, Antigravity (Gemini), OpenAI Codex, OpenCode (OpenRoute…
A developer built an OpenAI-compatible FastAPI reverse proxy that combines Redis Vector Search semantic caching with OpenTelemetry instrumentation, reporting a 58% reduction in LLM inference bills. The proxy matches prom…
OpenAI has produced proofs of long-open mathematics problems, including a result establishing that the chromatic number of the plane is at least 5 and a proof of the Erdős-Turán conjecture for arbitrarily long arithmetic…
A Kaggle benchmark called Reasoning Dial found that raising the reasoning_effort parameter from none to high produced exactly one statistically real accuracy gain across 1,800 graded calls on 4 models: gpt-5.4-mini on lo…
A developer built Cross-View Conservation, a Kaggle benchmark that tests whether language models can keep two representations of the same answer consistent as workloads grow, using deterministically generated synthetic s…
As of 11 October 2026, teams are replacing line-by-line code review of agent-written pull requests with six verification layers — AI review agents, risk routing, mutation testing, independent oracles such as property tes…
A developer connected a custom ChatGPT "dots" assistant to Gmail, Google Calendar, Google Drive, GitHub, Slack, Notion, Jira and Outlook and ran 17 real tasks over three days, finding that read-and-summarize work was lar…
A developer built SafeAir Walk, an open-source web app that identifies the single best 30–60 minute window to go outside each day by combining hourly air-quality and apparent-temperature forecasts from the keyless Open-M…
A developer published cache_gap_check.py, a read-only Python 3 stdlib script that scans Claude Code transcripts (~/.claude/projects/**/*.jsonl) for subagent sidechain requests, measures idle gaps between each subagent's …
A developer built a prompt-injection benchmark that planted a malicious AGENTS.md file in a repository to test whether frontier AI coding agents would blindly follow repo setup instructions. Both GPT-6-sol and Claude Opu…
Tej Pandya, founder of GrowEasy.ai, published a design sketch arguing that sales reminder agents should be modeled as explicit state machines rather than relying on conversation history as the record of what an agent owe…
India's Computer Emergency Response Team (CERT-In) issued critical advisory CIAD-2026-0046 on 9 October 2026 with CSIRT-Fin, warning that attackers are hijacking payment and disbursement APIs at NBFCs, digital lenders, f…
A developer benchmarked 10 AI models as judges on Kaggle, planting errors in their own prior answers to test self-preference, and found that only two small OpenAI models graded their own mistakes more leniently. The stud…
Ting-Wei "Willy" Sun, a 44-year-old contractor, pleaded guilty in Manhattan federal court on October 8 to four felony counts — conspiracy to violate US export controls, conspiracy to smuggle goods, conspiracy to defraud …
A developer measured Claude Code 2.1.289's `keep-coding-instructions` output-style flag by capturing the API requests it sends, finding the flag controls a 1,044-token "# Doing tasks" section that appears only in the ful…
Amazon Web Services will stop requiring government officials to sign nondisclosure agreements on new data center projects, AWS CEO Matt Garman announced in an October 2 blog post that also pledged more than $1 billion ov…
A developer writing as Mark Vange of Autom8ly re-ran the open-source decision model Kev-4B (4-bit quantized) over 866 questions and found that answers are bit-identical when requests run alone or in batches of two to thr…