We need an education moonshot

The perplexing challenge of education reform in the AI era: the Brookings Institute just shared the results of their year-long global study across 50 countries about the risks and opportunities of AI for students. It’s super dense and over 200 pages, but their conclusion is simply that the risks of using generative AI in education outweigh the benefits.

Their concern is that AI over-reliance could undermine children's foundational development - impacting their ability to learn, emotional well-being, relationships with teachers and peers, plus their safety and privacy. So... not nothing. 😅

They weighed these risks against the potential benefits: well-designed AI can enhance learning when implemented thoughtfully, expand access for neurodiverse or disadvantaged students, and enrich education when it deepens and expands the capabilities of - and interactions between - students, teachers, and content.

They offer 12 broad recommendations such as to use AI tools that teach, don't tell, promote holistic AI literacy, and establish comprehensive regulatory frameworks for educational AI. You can see the full report here.

My take: Nothing here surprises me - these risks echo what I've been sharing. But jeez, their conclusion is frankly demotivating. 😩 And, the recommendations feel somewhat tactical and old-school when we desperately need an inspiring, actionable vision for education that:

• Reimagines how we successfully educate kids for the AI era

• Makes learning personalized, engaging, and social - not about task completion and grades

• Prioritizes process, agency, and growth mindset

• Uses unhackable age-gating, deeply thoughtful product design, and straightforward regulation to ensure AI enhances - not harms - kids' foundational development, whether in classrooms or anywhere else.

In the meantime, my suspicion is that parents are going to need to do more heavy-lifting to support our kids as they navigate these AI risks and benefits.

PSA: watch out for Grok

This has gotten a lot of news coverage over the last week or so, but Elon Musk’s Grok chatbot recently rolled out an “Edit Image” tool that lets any user instantly modify others’ photos without the original poster’s consent or notification. Ever since, there has been a flood of nonconsensual sexualized AI edits including undressing women and, horrifically, children. Worse, Grok is embedded inside X, so these deepfake images have instant mass distribution.

Even more concerning has been Xai’s response, ranging from flippant acknowledgement, to canned responses, to blaming users for any illegal edits or CSAM. On Wednesday they finally said they would block this ability in countries where it is “illegal,” but for some reason that doesn’t comfort me too much.

An article in the Atlantic yesterday raises alarm bells about Musk seemingly getting away with allowing revenge porn and child sexual abuse to run rampant and the broader, unacceptable implications: “If there is no red line around AI-generated sex abuse, then no line exists.”

To do:

1. Audit the photos of your children that are published online. As a quick test, search for their name and see what comes up, but also check school and activity websites and your social media, especially if your accounts are public.

2. Have an age-appropriate conversation with your kids about AI and deepfakes:

   • If you post a photo or a video, someone can use AI to change it. It’s not your fault, but it happens.

   • If someone sends you one of these images, or if it happens to you, don’t forward it, don’t save it. Come to me and I will help you. You will not be in trouble.

   • If someone asks you to change someone else’s photo, that’s not ok, even if it is a joke.

Soft partying and friendfluence

There are some intriguing new social trends among Gen Zers that I thought were worth sharing, both to give us a boost of optimism for the social lives of our teens and young adults and because they feel applicable beyond Gen Z. 💪

1. Soft partying is a new and increasingly popular style of socializing that prioritizes daytime or early-evening events, low or no alcohol, and IRL connection and well-being. Think coffee house DJ sets (up +478% according to eventbrite), morning raves, and sauna and cold plunge sessions (up +246%). For someone who loves to dance but can’t handle the clubbing scene anymore, this sounds…. awesome! (and morning dance parties are up +373% here in Denver!). You’ll know where to find me…

2. Friendfluence is apparently one of the top 2026 dating trends, and a promising one (in moderation), IMHO. Think dating less as a lonely swiping solo mission and more like a team sport: double dates (Tinder’s new double-date feature is growing like gangbusters), group hangs, consulting your friend chat group, etc.

I love that both of these trends are heading in a promising direction whereby young people are prioritizing IRL connection and friendship while technology simply plays a very helpful facilitation role. Go Gen Z!

Personal Intelligence: Google plays their trump card

For a couple years now I’ve been wondering why Google doesn’t do more to connect Gemini to all the other products we use on a daily basis in a way that makes using their AI remarkably more personalized and useful, given how much history, data, and content it has about each of us.

To date they’ve been going product-by-product, integrating Gemini in little useful ways, but this week they announced the bigger swing, what they are calling “Personal Intelligence.”

The feature is opt-in and you can choose which Google services you want to connect (and can disconnect at any time), but this has the potential to unlock a lot more utility:

• Based on all my travel photos and trips I’ve taken, what would be my perfect next vacation - location, hotels, activities?

• Recommend some documentaries based on what I’ve been curious about lately and my interests more generally.

It’s rolling out this week to Gemini Pro and Ultra users, you’ll find a “Personal Intelligence” option in your Settings when it’s available. It will supposedly expand to free tier and AI Mode in Search soon. For now, it’s only available for personal accounts, not work accounts.

Just repeat yourself for better results

The “Say it Twice” hack: forget time-consuming, complex, elaborate prompts. New research shows that simply copying and pasting your prompt so it appears twice can boost accuracy by up to 76%. Because AI reads left-to-right, the repetition lets it “see” context it might have missed the first pass. It’s a free, zero-effort way to get better results instantly. Go figure.

Sweet dreams

To date, many sleep medications surpress REM as a side effect, but there is a new drug, BMB-101, still in clinical trials, that has reportedly boosted REM sleep by 90% in epilepsy patients without making them sleep longer.

BMB-101 could potentially have major impact in addressing cognitive decline, Parkinson's, and Alzheimer's by restoring one of the brain's most restorative sleep phases.

I’ll take one room on the moon, please

Well, looks like we can put deposits on just about anything now - if we have the $.

A YC-backed startup called GRU Space says it wants to build the first lunar hotel, taking deposits of $250K–$1M from wealthy true believers who are willing to reserve a bed on the moon years before the infrastructure exists. Their 22-year-old founder, Skyler Chan says they plan to launch their first moon hotel in 2032, with 4 rooms to start.