Full-text search across 70902 articles. Combine with topic and date filters; results sorted by relevance.
MCP released a new feature embedding product analytics and evaluation directly into agent sessions, enabling developers to capture response latency, success rates, and user interactions without manual logging. The lightw…
A Hacker News user asks why people who dislike AI continue to pay frontier labs for coding tools, and suggests using OpenCode to switch between models and avoid lock-in, enabling a future transition to open source. The u…
Knowl, an open-source agent memory system, retires stale facts when they change by splitting knowledge into atomic units and flagging conflicts as superseded. In benchmarks on MemoryAgentBench FactConsolidation single-ho…
A Hacker News user questioned the value of herdr, noting that harnesses like pi already ship most of its features, including agent multiplexing per tab, which is implemented in claude code, codex, and pi with plugins. Th…
Evals are standardized tests used to measure AI system performance, requiring unseen questions, scored answers, multiple test types, and comparisons across model versions to track trends over time. The piece explains tha…
Linus Torvalds credited an AI assistant for helping debug a Linux kernel issue that required 24 patches and 18 kernel boots to resolve, with the final fix being a one-line change from round_up() to round_down(). In a com…
LeadAce, an outbound sales agent built as a Claude Code plugin by a small Tokyo software company, launched on Product Hunt on August 19 and received zero signups and only two upvotes. The company's analysis revealed that…
AWS released aws-bench, an open-source benchmark to evaluate AI agents on real AWS tasks, using disposable AWS accounts and automated verifiers. The benchmark, built on Harbor, supports agents like Claude Code, Codex, Ki…
A developer warns that silent model upgrades can break AI agents without any code changes, as behaviors like answer length, tool choice, and ambiguity handling shift. The post recommends building a frozen baseline of rea…
Ram launched Anjadhe, a privacy-first AI assistant for macOS that runs locally with llama.cpp, stores data in SQLite, and requires no account or server database. The early alpha app reads email to create tasks, manages g…
Postcard, a new email client for GNOME built in Python with GTK4/libadwaita, was released by solo developer gxanshu on GitHub, who used AI coding tools to create the app. The client features a Geary-like multi-column des…
Ox Alpha, a stealth model available on OpenRouter, ranks #4 on the Elo Rating in the Multi-Agent Arena, just behind GPT-5.6 Sol, and is the first frontier model not from OpenAI or Anthropic. It features a 1M token contex…
Fortitude Omnis Group's OmnisBench benchmark initially showed small models performing surprisingly well, but community feedback revealed potential contamination from old benchmarks like HumanEval and GSM8K. Upon testing …
Vercel's new agent-readiness test, is-agentic.com, scored 67 for a production docs site, failing three Essential checks that block headless agents: empty HTML shells on dynamic routes, 200 status codes on error pages, an…
PAPI: Persistent Adaptive Project Intelligence, an adaptive plan/build/review cycle tool for AI coding assistants, was published as version 0.7.85 to the official MCP registry on 2026-08-21, distributed as @papi-ai/serve…
A developer built mcp-anything, a meta-server that indexes and searches over 75,000 MCP servers, providing five meta-tools for discovery and execution while keeping context costs constant. The project uses BM25 lexical s…
Alibaba and Traccia have launched competing observability platforms for production AI agent monitoring: Alibaba's AgentSight, an eBPF-based tool requiring no code changes, and Traccia, which offers runtime control and au…
A developer trained a 125M-parameter transformer for real-time MIDI autocomplete, achieving approximately 108 notes per second on an iPhone 15, modeled after GitHub Copilot's code completion approach but for piano perfor…
The CodeYam team released a second video showing early design explorations for CEO and cofounder Nadia Eldeib's personal website redesign, using CodeYam with Claude Code as the AI agent to iterate on design approaches ac…
Google has won a $10 million bid for Spirit Airlines' corporate data, including internal wikis, emails, source code, and spreadsheets, to train AI models, according to court documents. Rival AI training provider Micro1 h…