Full-text search across 26032 articles. Combine with topic and date filters; results sorted by relevance.
Google released Gemini 4 Argon on 30 September and moved it to general availability, offering a 1,048,576-token context window, text/image/video/audio/PDF input, and introductory API pricing of $2 per million input token…
A security researcher traced more than 16,500 scans of the UN Conference on Trade and Development's statistics platform (UNCTADstat) to OpenAI agents active between April and June 2026, based on session labels such as "C…
A developer building Qwellar, a desktop app that gives local AI models persistent memory via Ollama, reports that Ollama's stock API defaults — temperature 0.8 and num_ctx 2048 — degrade the output of small local models …
Meta launched Muse Spark Open on Thursday, an open-weight AI model that Meta's chief of AI Alexandr Wang called "the most capable US open-weight model to date" and that Meta claims rivals or outperforms leading Chinese o…
Nango tested tool-exposure patterns for AI agents and found that exposing a small pinned working set with on-demand discovery cut median cumulative input tokens per task by 69% versus loading all 57 tool definitions upfr…
Google DeepMind's Gemini 4 Argon safety report found no statistically significant acceleration in AI-driven machine learning R&D, with its internal GCI v1 capability index rising a steady ~0.1 points per month, researche…
A developer published a guide on getting better planning results from Claude Opus and Sonnet, arguing that model choice matters less than giving the model live schedule data, short constraint-focused prompts, and persist…
Founders presenting AI agent spend to boards should report per-task unit costs rather than a single lump monthly figure, according to guidance citing Cursor's engineering team's published estimate of roughly $20 to $40 i…
A team auditing its own browser-based AI agent harness identified 21 gaps across 9 failure modes and fixed all 21 in a single epic with 36 sub-issues, merging with 8,047 tests passing. The audit found the failures were r…
The Model Context Protocol's 2026-07-28 specification requires that a human always be able to deny a tool call, that `tools/list` return only the tools available to the requesting client, and that tool annotations be tre…
Playgama launched a suite of tools — Playgama MCP, Playgama Sandbox, and the Playgama Bridge SDK — that lets AI agents publish, analyze, and update HTML5 games autonomously, founder Dmitry Kachmar said in a press release…
Scuba, a New York-based startup founded by former Slack chief product officer Noah Desai Weiss and engineering leader Dolapo Falola, launched publicly with a "context engine" that gives teams and AI agents a shared memor…
Workday, Inc. (NASDAQ: WDAY) announced on Oct. 13, 2026 at Workday Rising 2026 that its Trusted Agent Platform is now headless, letting customers build, run, and govern AI agents across HR and finance using their own dev…
TabularisDB released v0.28.0, a database client update with more than 80 commits and close to 70 merged pull requests from nineteen external contributors, arriving five days after v0.27.0. The release rebuilds the struct…
Tool output accounted for 56% of context tokens and tool calls another 28% in a 22 September 2026 CMU study of coding agents on Terminal-Bench 2.0, according to the CliffCompaction paper from Tim Dettmers' group. The stu…
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 …
Google's own benchmark table for Gemini 4 Argon shows the model taking the top score in 13 of 19 rows, with GPT-6 Astra winning 3 and Claude Opus 5.5 winning 2, according to figures Google published on its Gemini 4 Argon…
A developer who builds the paid Mac app Claude Sessions Pro documented how to run multiple Claude accounts simultaneously by giving each account its own data directory: setting CLAUDE_CONFIG_DIR for Claude Code CLI sessi…
A freelance software engineer built ReceiptDesk AI, a privacy-first invoice and receipt auditor that runs entirely on-device and is live in production at shivysidinfo.tech. The tool uses Google's open-weight Gemma 2 mode…
A developer built hatswap, an open-source CLI that keeps multiple Claude Code accounts signed in simultaneously by giving each account its own CLAUDE_CONFIG_DIR under ~/.claude-accounts/<name>, while symlinking settings,…