Well, so much for being able to give last week’s “Hands-on with AI” demo a try (Anthropic’s groundbreaking Claude Fable model) … because just a day later, Anthropic was forced to remove access to the model. The US government demanded they deny access of it to any foreign national (including those living here in the US and even many of Anthropic’s own employees).

The stated reason was that someone found a “jailbreak” - a way around the model's safeguards - that the government is worried could be used to pull out information useful for a cyberattack. On the surface I would think - great, I am glad we are being safe and smart. (And, hey, aren’t all the AI leaders asking for frontier AI regulation?)

But under the surface things get pretty suspect. The vulnerability they're “worried” about is not just a Fable problem, it's a frontier-model problem. Every lab at this capability level is wrestling with the same class of jailbreak. So singling out one company's models, while leaving the others running, doesn't obviously make anyone safer. Moreover, Anthropic states that the vulnerabilities are previously known, pretty simple, and relatively minor.

If you recall, Anthropic isn’t exactly on the best of terms with the US government. Here is Anthropic’s statement about the situation. Admittedly, I don’t know all the details nor what’s truly going on behind the scenes, but this seems less like an objective safety response and more like a vindictive political game - and a very dangerous precedent to me.

Google’s longtime CEO (and someone I deeply respect), Sundar Pichai, delivered the commencement address to Stanford and shared 3 filters he lives by. When I read his speech, they struck me as beautifully simple and deeply relevant for navigating today’s complexities:

As a parent, I love these 3 filters. They are teachable and tangible for any age: look for the bright side and believe in yourself. Say yes to the hard thing. Chase what excites you.

We don't need a Stanford diploma to start practicing these, and honestly, the practice is the whole point!

As a whole, we have never been physically healthier than we are today. Gym memberships are at record highs, running clubs quadrupled last year, drinking is at a record low, and wearables are giving us daily scoreboards to beat. Yet also, we are lonelier than ever. Young people report fewer friends and more time alone than ever measured.

Derek Thompson believes that these two facts are not coincidental, but connected: we have become so obsessed with self-optimizing that it is impacting our spontaneity and our social lives:

→ The best way to win the sleep score is to skip the dinner out with friends.

→ The best way to protect our midday workout is to cut lunch with friends or coworkers.

By it’s very nature, optimizing ourselves is a solitary project, and it can slowly crowd out the unquantifiable stuff that actually makes for a long and happy life.

So… how do we actually socialize more when life is already full?!

Another good article this week says it clearly: just have people over! Aka: stop reserving friendship for the clean house, the planned dinner party, the free Saturday that never comes.

Rather, invite people into the unglamorous thing you're already doing. Fold laundry together. Let them watch you make a mediocre weeknight dinner. Lift weights together in your basement. Lower the bar so that connection stops competing with everything else on your list and just slots into the life you're already living.

Try this: pick one ordinary activity this week - working out, waiting for your kids to finish swim practice, watching a new show on Netflix - and invite someone to do it with you instead of saving the invite for an occasion that requires cleaning the house first.

Let your kids see that connection isn’t about keeping score, it’s about keeping a low bar and an open door.

There’s a good chance you’ve heard about - and perhaps already use - Google’s NotebookLM, but I thought it was worth a feature this week because, a) this tool is one that a lot of schools are using/allowing for student study aids. And so, if your kiddo is using it, you should get to know it too! and b) they’ve recently introduced some new cool features.

What it is: NotebookLM is an AI product from Google that helps you learn and explore and manipulate specific content (that you upload or ask it to source for you) in a wide range of ways.

Give it a try: A fun way to give NotebookLM a try is to pick something relevant to your summer (e.g. the World Cup, where you’re heading on vacation, etc.) to learn more about. For me, the World Cup was a great one because I know so little about the players and the teams but want to ramp up quickly - and it’s a fun topic to learn about with my boys!