Summer is finally here ☀️ and  I've been doing what you're supposed to do in summer. Staying outside, swimming, seeing friends & family, and going to sleep way past my usual bedtime. I'm also leaving my phone inside whenever I can.

Which means I didn't really realize we just had two of the wildest weeks in AI this year. Three new models launched. One got banned by the US government three days in, then unbanned 19 days later. And Claude gave everyone a smarter free version without asking. All of it happened while I was on a pool float, probably sipping on a Brazilian lemonade 🍋.

Want to know what surprised me? Missing it cost me nothing and changed very little in my day-to-day. The AI I was using two weeks ago still works exactly as well today.

If you saw the headlines and thought "what does any of this even mean," this issue is for you. Let's break it down.

When people say "AI model," they mean the 🧠 inside the app. ChatGPT is the app. GPT-5.5 is one of the brains it can use. Claude is the app. Sonnet 5 and Fable 5 are brains it can use.

Now think about iPhones.

Every year, a new one is released. It usually has a better camera and a faster chip, and people line up to get it. Meanwhile, the phone in your pocket keeps taking great photos. It didn't get worse overnight because a new one exists.

AI models work the same way. Labs keep training new versions that are smarter, faster, or cheaper. But the one you were using yesterday is still very good at what you were using it for.

So "which AI is best?" is the wrong question. The right one is "which is right for what I'm doing today?" Here's my short version:

Sonnet 5 is the phone in your pocket. It's now what Claude gives everyone for free, and it handles almost everything: emails, summaries, planning, brainstorming.

Fable 5 is the pro model with the fancy camera. It's Claude's most powerful brain, on paid plans. Save it for the big stuff: long documents, tricky analysis, multi-step projects.

ChatGPT's models are a different brand of phone but essentially the same idea with different strengths. If you're already comfortable there, you're not behind.

And if you use Fable 5 when Sonnet would have done the job? Nothing breaks. Your answer is fine. It just arrives slower, and it burns through your usage limit way faster. Sometimes those big, more powerful models overthink things. You asked for a two-line email and got a strategy memo.

So the real risk isn't picking wrong. It's paying in time and usage for power you didn't need. The model matters less than what you type into it. Which brings me to tonight.

There's a science name for this: the protégé effect. The person doing the explaining learns the most. So instead of asking your AI to explain something to you, you're going to explain something to it. Copy this into Claude, ChatGPT, Gemini, whichever one you have open:

Start with the topic you just read: what an AI model is. Explain it using the iPhone thing and out loud if you can.

You read this newsletter, you teach it back to an AI tonight, and on Saturday you're the person at the BBQ explaining models with an iPhone in one hand and a burger in the other. Sounding smart isn't a trick, it just takes a little practice.

Want to know which model to open for which task? I made you a one-page cheat sheet.

🪁 Kids are already asking AI for life advice. A new UN report says millions of children around the world are using AI for homework help, and increasingly for personal worries too. The safeguards? Not keeping up with how fast kids have adopted it.

Why it matters: if there's a kid in your life with a phone, this is probably already happening. You don't need to panic, you need to be someone they can talk to about it. Knowing the basics (like what a model is, hi, this issue) is how you stay in that conversation. → Read it here

🏛️The US government grounded an AI. For 19 days. Anthropic launched Fable 5, its most powerful model yet, on June 9. Three days later, the government pulled it offline over security questions. Testing later showed other AI models could do the same things it was flagged for, and it came back on July 1.

Why it matters: the tools we use can change overnight, and that's one more reason not to marry a model name. Know what you use AI for, and you can do it anywhere. →Read it here

⚖️ Illinois just became the strictest state in the US on AI. Governor Pritzker signed a law requiring big AI companies to get audited every year by an outside safety expert, first state to require that. They also have to report serious safety incidents within 72 hours. Anthropic and OpenAI both supported it.

Why it matters: this is the third state to pass a law like this. Between Illinois, California, and New York, that's about 40% of the AI market now under similar rules. →Read it here

But here's an idea for your next BBQ: make AI the topic. Share the one trick you know. Ask what everyone else uses theirs for. Someone at that table has a use you've never thought of.

It's called the production effect: you remember things far better when you say them out loud than when you just read them. Add the protégé effect from tonight's exercise, where the one teaching learns the most, and that BBQ conversation stops being small talk and turns into a strategy.

What's the one AI trick you'd share at a BBQ? 🍗🍹

À la prochaine,

Cinz

Cinz.AI

Your unfair advantage, one week at a time.

Got a question, a topic request, or just want to say hi? Just reply to this email. I read everyone.