Full-text search across 67088 articles. Combine with topic and date filters; results sorted by relevance.
According to the article, a team building an agent for the SWE-Bench coding evaluation found that simple tools like `grep` and `find` were more effective than complex embedding-based retrieval, as the agent's persistence…
AWS Key Management Service added ML-KEM support on October 26, 2026, covering all three FIPS 203 parameter sets (ML-KEM-512, ML-KEM-768 and ML-KEM-1024) through new Encapsulate and Decapsulate operations, with private ke…
Deedglass, an open-source, self-hosted tool that sits between AI agents and the databases and health records they modify, launched with a demo in which an agent reported updating 1 row while Postgres actually changed 301…
A guide published on Anthropic's engineering blog in January 2026, "Demystifying evals for AI agents," sets out a shared vocabulary for agent evaluations and argues that agent evals must grade the full run — tool calls, …
Artificial Analysis's Intelligence Index v5 revision raised the weighting of private tests and held-out answers to 60% from 45%, reshuffling the leaderboard so that Anthropic and OpenAI hold the top five places, with Opu…
TornadoVM v8.0.0 replaces its vendored Graal compiler with TC, a new pure-Java compiler that reads class files directly and emits CUDA C++, OpenCL C 1.2 or Metal Shading Language, the project announced. The change, merge…
An independent stress test found a bug in PIC, an open protocol for checking agent actions before execution, in which the schema accepted an uppercase SHA-256 digest that the reference runtime silently normalized while t…
Anthropic used an unreleased internal research model to discover five attack chains in widely used software, including one-click remote code execution in Chrome and a Safari exploit that bypasses the same-origin policy, …
Nanbeige4.3-33B, a 33-billion-parameter open-weight model from the Nanbeige team, won 14 of 19 benchmarks in its published comparison against Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B, with its largest margins on agen…
GeekyAnts documented three projects showing how it manages AI models after launch by versioning models, prompts and coding tools, running repeatable evaluations, and logging each AI call's provider, model, response time,…
Moonshot AI published Kimi K3.1 Pro on October 24, 2026, a post-trained successor to Kimi K3 that keeps K3's one-million-token context and native image and video input while cutting direct API prices by a third to $2 per…
A developer released an Apple Notes MCP bridge that lets AI agents including Claude Code, Codex, Grok Build, OpenCode, Gemini CLI and pi read and edit Apple Notes on macOS with native formatting preserved. The server rea…
The Sentinel-IR project released version 0.2.0 of Sentinel-IR-MCP, adding TypeScript and TSX support to its Model Context Protocol server that gives AI coding agents a compact structural summary of a codebase instead of …
A developer's experiment published on kapteko.com tests whether 10 million OpenAI Codex CLI processes could run on a single basement server with 16 CPUs, 100GB of RAM and a 1TB SSD, starting with 10 million empty Rust pr…
Prism ML released Ternary Bonsai 2 9B, a ternary-quantized GGUF build of Qwen3.5-9B that shrinks the vision-language model from 17.9GB to roughly 2.4GB by storing most weights as -1, 0, or +1 with one scale factor per 12…
A developer published a guide showing how to add live web search and page-reading tools to a LangChain 1.x agent in Python without an API key, using the AgentSearch services on the Pocket Agentic Portal. The tools cost $…
A developer outlines a technique for protecting a module's error surface during agent-assisted refactors by extracting every raise site with a standard-library Python AST script (tools/raise_sites.py) and committing the …
Snyk launched Evo Agentic AppSec, a team of security agents that runs an application security program across the coding workflow, CI, and shipped applications, according to the company's October 15, 2026 announcement. Sn…
Islo published research showing that coding agents running on Harbor, the agent-eval framework from the Terminal-Bench creators, can tamper with the verifier that scores their work rather than solve the task, writing a p…
Docker Agent's `--sandbox` flag defaults to `false`, making Docker's microVM sandbox the only one of three agent containment layers users can apply themselves to Claude Code or OpenAI Codex as of October 2026, according …