Latest
Four Controls That Turn an AI Agent From Demo to System
Two sponsored 2026 agent guides, product names stripped out, leave the same four controls — and four questions you can make any vendor demonstrate live.
How to Get Your Brand Cited by ChatGPT and Perplexity
AI assistants name two or three brands per answer. A 30-minute manual audit, the factors driving citations, and the llms.txt rules three new guides prescribe.
Microsoft: Vulnerabilities Are Weaponized in Under 24 Hours
Coverage of Microsoft's 2026 Digital Defense Report puts median weaponization under 24 hours, phishing at 23% of intrusions, and 32-step autonomous attacks.
Cloudflare Routes Worker Errors Straight to Your Coding Agent
Issues for Workers hit open beta September 30. It hands exceptions and the affected Worker version to a coding agent — triage them before you let it patch.
OpenAI Safety Report Author Quits Over 'Broken' Culture
David Robinson, who wrote OpenAI's launch safety reports for three and a half years, says frontier labs need nuclear-plant discipline. OpenAI responded.
Apple Tightens macOS Full Disk Access Over AI Agent Risk
Apple will add new controls to macOS Full Disk Access, warning that increasingly autonomous AI agents make system-wide file access far riskier.
LMCP Puts 190+ Mac App Tools Into Claude Code for Free
A free Mac app exposes Mail, Teams, Slack and browser tools to Claude Code by reading local app caches instead of cloud APIs. The vendor's limits matter.
OpenAI's Dots Agent Costs $100 a Month and Still Needs You
OpenAI gates its Dots agent behind a $100-per-month tier while Muse and Instinct stay free. Hands-on results split, and a patched macOS flaw raises the stakes.
GPT-6 Astra Ultrafast Is 8x Faster. The Price Isn't Public.
OpenAI's GPT-6 Astra Ultrafast runs on NVIDIA Blackwell and claims up to 8x faster token generation, but the launch announcement names no price at all.
Cloudflare Opens a 402 Beta: Agents Pay Per Call in USDC
Monetization Gateway prices websites, APIs, MCP tools and datasets per request over HTTP 402, settling in USDC on Base. Four customers are live, US only.
Claude Opus 5.5 vs GPT-6.1 Sol: Effort Flips the Winner
Opus 5.5 wins at max effort, GPT-6.1 Sol at medium, and Sol costs less per task in every published row. List prices, a cost-per-turn model, and the catch.
A 400MB Model Ran a Browser Agent on a 2017 Galaxy Note 8
Qwen3-0.6B scored 10/10 on a live browser task from a 2017 Note 8 and 0/3 without a perception layer — all figures self-reported by the layer's team.
Restate Raises $20M to Take On Temporal in Agent Durability
Berlin's Restate raised a $20M Series A led by Singular. It was built for durable workflows in 2022, not agents - and that turned out to be the fit.
