Updated weekly ·

This Week in AI

A weekly summary of what changed in AI: new models, agents, and workflows from OpenAI, Google, Anthropic, and Microsoft, plus what it means for your role. Every update links to the official source so you can check the details yourself.

See this week's updates Subscribe via RSS Get the weekly email

What changed this week

Each item links to the official source and one thing you can do about it.

Models Official source

Thinking modes are the default for hard tasks

The main model families let you turn up reasoning effort or extended thinking. Instead of memorizing version numbers, check the live model pages: OpenAI, Gemini, and Claude.

What to do: Learn when to switch to thinking mode with our reasoning models guide.

AI Model & Agent Tracker

The current landscape at a glance. Each row links to the vendor's live model page so you can confirm the exact current version.

Vendor Flagship assistant Agent / build platform Live model page
OpenAI ChatGPT (GPT-5 family, thinking) AgentKit · Connector Registry platform.openai.com/docs/models
Google Gemini (with Deep Think) Vertex AI Agent Builder · Gemini Enterprise ai.google.dev/gemini-api/docs/models
Anthropic Claude (Opus/Sonnet/Haiku, extended thinking) Tool use · MCP · computer use docs.anthropic.com/…/models
Microsoft Copilot (in Microsoft 365) Copilot Studio · Agent 365 · Work IQ learn.microsoft.com/copilot

Version names change fast. Always confirm the current model on the vendor's live page linked above. See our full tools comparison →

Prompt of the week

"Act as my reviewer. Here is a draft [memo/analysis]. Challenge every claim that isn't supported by the source I pasted, list the 3 riskiest assumptions, and tell me exactly what a human must verify before this ships."

Browse the full Prompt Library →

Never miss a week

Get the weekly AI brief in your inbox: new models, agents, and one action to take. Prefer a reader? Subscribe via RSS. No spam, unsubscribe anytime.