Gemini 4 Carbon is GA, and the $2 price has an expiry date
Gemini 4 Carbon is generally available with a /swarm preview in Antigravity. What the benchmarks and the price change on 30 November mean for your team.
Two rubrics, a new post every other day: the AI and software news worth your time, and teardowns of the AI technologies everyone is talking about.
Gemini 4 Carbon is generally available with a /swarm preview in Antigravity. What the benchmarks and the price change on 30 November mean for your team.
Claude Agent SDK packages the Claude Code agent loop for Python and TypeScript. We cover what it does well, where it bites, and who should build on it.
GPT Image 2.6 (Sunburst and Flare) adds SVG and GIF output and a complexity setting. See what's confirmed, what's a vendor claim, and what to test this week.
CrewAI has the biggest community of open-source agent frameworks. We look at Crews, Flows, durability gaps and who should use it in production.
GitHub Copilot Fusion is GA with Standard, High and Max effort levels. What the routing does, what the cost claims cover, and what to test this week.
LangGraph 1.2 in plain terms: how graphs, checkpoints and interrupt() work, where the verbosity bites, and who should pick it over CrewAI or OpenAI's SDK.
GPT-6.1 Terra costs $1 in and $5 out per million tokens. What OpenAI actually claims, what is unverified, and what to test before you reroute agents.
What the OpenAI Agents SDK does well, where it costs more than it looks, how it compares with LangGraph and the Responses API, and who should use it.
Microsoft Agent Framework hit 1.0 in April 2026. We tear down how it works, where it's strong, where it costs more than it looks, and who should skip it.
LiteLLM 1.107.0 stole cloud and AI keys through Python startup hooks. What Microsoft found, what's still unclear, and what to check on your hosts this week.
Anthropic put Claude on Google Cloud Marketplace on October 22. Here's what changes for billing, governance and org setup, and what to do this week.