Full-text search across 68692 articles. Combine with topic and date filters; results sorted by relevance.
A four-person team including Max Kanat-Alexander of Capital One, Eirini Kalliamvakou of DX, and Margaret-Anne Storey of the University of Victoria introduced the CAFE(S) framework on the Engineering Enablement podcast, d…
A developer has ported raylib 6.2 to Zig, checking every example against the original C version and running the library in the browser via WebAssembly. The port is a direct C-to-Zig translation that calls libc functions …
The Unified Harness Protocol (UHP) was published as a draft open standard, version 2026-10-04, defining a universal translation layer between agent harnesses such as Codex, Claude Code and Hermes and the applications tha…
Pensieve's benchmark, published this week, found that Claude Code averaged 3.4 tool calls per file lookup using grep, glob and read across 72 tasks in four real repositories, versus 2.0 with Pensieve's local index attach…
A new arXiv benchmark, Into the Unknown, tasked nine frontier AI agents with writing 15 follow-up papers each on unpublished research directions contributed by scientists at US Department of Energy national laboratories,…
Simon Willison released Chirp Pong, a two-player pong-style game that runs as static HTML on two separate phones and transmits game state through audio chirps picked up by the web microphone API. Willison built the exper…
Federal prosecutors unsealed a criminal complaint in San Francisco charging OpenAI product engineer Abdulrahman Alfozan with intercepting oral communications after a $23.99 Seeed Studio XIAO ESP32-S3 Sense recorder was f…
Researchers from Google Cloud AI Research, Google DeepMind, and Carnegie Mellon University introduced Harmony Harness, a framework that treats large language model agent scaffolding as a multi-objective optimization prob…
Epoch AI's Capabilities Index, updated October 1, shows Claude Opus 5.5 leading at 167.35, with GPT-6 Astra at 166.51, Claude Sonnet 5.5 at 165.2 and Claude Fable 5.1 at 164.82, while Google's Gemini 4 Argon does not app…
An open notebook maintained by MV2 reports a new bound on the growth base of rank-r Kakeya sets over a fixed finite field, showing beta(q,r) = q - q^(2-r) + O(q^(1-r) log q) and giving constructions smaller than the 2010…
A developer built Order Ledger, an offline, MIT-licensed tool that converts messy WhatsApp order chats in English, Hinglish and Kannada into structured order records using a local open-weight model. The app runs entirely…
Vercel's $1 million sandbox escape challenge drew 1,285 vulnerability reports and paid out $242,000 in bounties, with the top finding a Linux kernel networking flaw that let a microVM guest leak host memory, crash the ho…
Anthropic replaced Git with a custom centralized version control system called Scion after its commit rate jumped from 2,500 to 30,000 per day following internal deployment of Fable and Mythos-class models. Built by a se…
Rebuilding an enterprise AI platform with AI cut years of engineering work to a few months but exposed an "architecture tax" that drove token consumption far beyond what any model upgrade could offset, according to the c…
Simon Willison released a LaTeX renderer at tools.simonwillison.net/latex that stores LaTeX source in the URL fragment, letting rendered documents be bookmarked and shared as links. Willison built the tool with Claude Co…
Wes McKinney's kenn.io released Robot, a local multi-agent development control plane that coordinates work across four tools plus Git: Kata for issue and work tracking, kwt for git worktrees and tmux workspaces, Roborev …
AWFlow's AI agents run on deterministic workflows rather than autonomous agent loops, a design the company says makes runs repeatable, debuggable and cheaper, according to a company post explaining the choice. AWFlow sta…
Under the 2026-07-28 Model Context Protocol specification, the three MCP primitives — tools, resources and prompts — differ mainly in who decides when they are used: the model calls a tool, the application attaches a res…
Pluto Security CEO Shahar Bahat said 60% of the risks his company detects come from employee-built "vibe-coded" AI agents, with finance departments the primary source, followed by marketing and legal, and developers rank…
Meta's new frontier models Muse Wonder and Muse Wonder Full ranked first on three agentic tasks in the How I AI bench, a roughly 20-task evaluation spanning prototyping, agentic workflows, coding, writing, and creative w…