AI in a Nutshell - Week 35 - Microsoft Goes Solo, Claude Wants Your Chats, and Meta Borrows Google’s Muscles
Microsoft ships its own models, OpenAI+Anthropic do a rare safety collab, Anthropic is coming to Chrome, Meta rents Cloud from the ops (Google), Intel gets a government stake, and xAI drops a speedy c
Fellow human, it’s week 35. AI hasn’t taken over the world yet. And here’s your lazy-man version of what went down in the AI world in the past week.
Speed-read: Microsoft ships its own models, OpenAI+Anthropic do a rare safety collab, Anthropic is coming to Chrome, Meta rents Cloud from the ops (Google), Intel gets a government stake, and xAI drops a speedy coder.
What You Must Know
Gemini’s glow-up for photos → Google juiced image editing in Gemini: outfit swaps, style transfers, photo blending. It’s called nano banana, and everyone is going bananas over it.
xAI releases Grok Code Fast 1 → A lean, fast, agentic coding model for shipping code with less faff. It’s been gaining attention on the internet.
OpenAI × Anthropic do a joint safety eval (yes, together) → The rivals teamed up to test models for “suck-up tendencies” (over-agreeing) and risky behavior. The aim is shared standards so models stop acting like overeager interns. Love to see it.
Meta patches teen-safety rules after watchdogs pounce → After reports of bots flirting with teens, Meta banned romantic chats with minors and tightened policy.
Meta signs a 6-year, ~$10B Google Cloud deal → Enemies? Partners? Yes. Meta is renting Google’s compute to scale AI. Moral: even giants need bigger gyms.
Microsoft’s “we have our own now” moment → Microsoft unveiled in-house models (MAI-1 and friends) trained across a mountain of GPUs. They can now lean less on OpenAI.
What’s Good to Know
TIME Magazine dropped the 100 Most Influential in AI → The usual suspects (Altman, Musk, Huang) appear alongside new faces (e.g., DeepSeek’s Liang Wenfeng). Guess what? Your name is not there.
Claude for Chrome (early access) → Anthropic is quietly testing a Chrome experience with premium users. It’s in its early days, but has an interesting direction.
Viral Roast of Google’s AI Overviews → Twitter community roasted Google’s AI Overviews for basically breaking Google’s own spam rules (ironically). The internet remains undefeated.
Anthropic settles the author copyright suit → A mega-damages risk disappears with a settlement. It’s not the last copyright story, but it keeps Claude’s roadmap intact.
Intel gets a fresh 10% government stake → The U.S. poured billions into Intel to keep cutting-edge chipmaking on home soil.
AI Tools Worth Knowing
Wanderboat - Social + local + AI map search from ex-Bing folks.
Surf (AskSurf) - AI-powered crypto insights + trading in one place.
Jeeva AI - Find verified emails and phone numbers.
Mocke - Mock entire email campaigns with AI.
Stormy.ai - Agents for influencer marketing.
Clueso - Create product videos with AI in minutes.
Portia AI - Secure AI agents with auth + tool controls.
Kaizen Corner - What is MCP (Model Context Protocol)?
The short version: MCP is a common way for apps, tools, and LLMs to talk to each other safely and predictably about context: files, data, and functions.
What it does: It gives language model apps (like ChatGPT or Claude) a consistent “port” to hook into tools, files, databases, or any other external resource, without needing a custom connection each time.
Why you should care:
Plug-and-play agents → Swap tools without rewriting code.
Safer automations → It helps provide guardrails by default.
Faster builds → Less API connections, more shipping.
Think of MCP as USB for AI tools: one clean port instead of 50 awkward adapters.
Meme of the Week
Question → Your Take?
Would you let your chat history train future models if it meant noticeably better answers for you in a month, or would you opt out on principle?
A) Train on it. I don’t care. I want smarter replies.
B) Hard no. My chats are my chats.
C) Fine… but only if I get a toggle per thread.
That’s your week in AI. Short, sharp, and just enough for you to stay plugged in.
If you learned something, tell a friend. And if you didn’t, blame yourself.
Until next Sunday,
Kay - your fellow human
P.S. If this email lands in spam, that’s just your inbox trying to stop you from staying plugged in. Fix it.