Welcome Back to The Cwtch 🧡
A cozy corner of the internet where AI makes sense — no degree required.
This week: Why one company's ethics matters more than you think.
🛡 THIS WEEK'S STORY: Anthropic's Refusal
You probably heard about Mythos — Anthropic's new AI that's supposedly too dangerous to release. It goes a little deeper than that.
Here's what actually happened:
The Trump administration asked Anthropic to remove the guardrails that prevent AI from being weaponized. Translation: "We want to use this for military purposes without safety limits."
Anthropic said no.
The government responded by banning federal use of Anthropic's technology and calling it a "supply chain risk."
Then Mythos dropped. And suddenly everyone's talking about cyber threats instead of the refusal.
Why this matters to you:
Those guardrails didn't appear by accident. Someone chose to build them. That choice comes at a cost — including the cost of refusing people you wish you did not have to refuse.
Most people think AI safety is a tech problem. It's not. It's a power problem. Who decides what AI does? The builder, or the person with the stick?
Anthropic picked ethics. That decision matters.
🔧 THIS WEEK'S TOOL: GOOGLE GEMINI
Google Gemini is Google's AI assistant. It's built into Gmail, Google Docs, Google Search, and Android phones.
What it does: Answers questions, writes, summarizes, brainstorms, explains things. And it's connected to your Google stuff, so context flows naturally.
Who it's for: Anyone already using Gmail or Google Docs. Anyone who wants AI that's familiar and doesn't require learning a new interface.
Why it matters: Gemini is one of the few major AI tools that's independent from the recent power plays. Google built its own and ships it everywhere.
How to try it in 3 steps:
Go to gemini.google.com or open Gmail
Click the Gemini icon
Ask it something practical: "Summarize the emails in my inbox from this week and tell me what I should follow up on"
Gemini reads your actual emails and gives you a real summary. No new account. No learning curve. Just ask.
⚡ QUICK WIN: THE DECISION-FRAMING TRICK
When asking AI for advice, try this:
"I'm making a real decision here. Show me the tradeoffs — what works, what doesn't, and what I should watch out for."
AI defaults to "safe" answers. This prompt makes it think like someone with skin in the game.
🌍 CWTCH WATCH
Perplexity $1M startup competition opens registration April 14 (3 days). Up to $1M funding + $1M compute credits. Build a company in 8 weeks. Builders: mark your calendar.
Google Gemma 4 is live — open models designed for reasoning and autonomous workflows. Translation: agentic AI is now mainstream.
ChatGPT: 900M weekly users, 2B daily queries, $10B revenue. Market leader. But watch the edges — competition is real.
Search trust shift: 88% of AI search users don't fact-check results (vs. 56% in regular search). We're trusting AI faster than we should, maybe.
💡 THIS WEEK'S PICK
Claude Web Browsing (Free at claude.ai) — Ask it about anything happening right now. It searches the web to give you today's answer, not yesterday's training data.
No subscription needed to start.
ONE LAST THING
The Mythos story isn't really about whether the AI is dangerous.
It's about who controls the choices.
Every tool you use — ChatGPT, Claude, Gemini — exists because someone decided it should. When a company refuses pressure, that matters. Because it means guardrails are chosen, not inevitable.
Hit reply. Which AI do you use most? 🧡
The Cwtch is published weekly. Unsubscribe anytime — no hard feelings.
This newsletter may contain affiliate links. They cost you nothing.