In partnership with
My fellow AI explorers
Today, you can learn about Google’s roadmap for how you can have a real conversation with Google AI; no more taking turns. Now you can get things done in real time.
School's back in session, and so is the scramble to figure out what AI is actually allowed to do in the classroom. This week we're covering the patchwork of states writing the rulebook in real time, and a Pope who just told legislators exactly where he thinks the line is,
In today’s edition:
🎒 Schools are teaching AI literacy, and the map just proved why
⛪ Pope Leo XIV draws a line on AI and the family
🎙️ Build a live, talking, browser-controlling AI agent (30-Second AI Play)
30-Second AI Play
🎙️ Build an AI Agent That Talks, Sees, and Clicks, Live
Most AI agents you've used are text in, text out, one turn at a time. This one is different: it hears you talking, sees what's on the page, remembers context from earlier, talks back, and moves a real browser while you're still mid-sentence.
Google calls the concept an "omni-app," since it can perceive, reason, and express running as one continuous loop instead of a back-and-forth conversation. Annie Wang lays out the full six-episode build in this roadmap preview using the Gemini Live API and Google's Agent Development Kit.
Here's how to actually work through it:
Watch the roadmap preview first to understand the core idea: perceive, reason, express as one live loop, not a turn-based chatbot.
Start with Episode 1, Voice, to see why a live session beats basic text-to-speech, and why letting the user interrupt changes the entire interface design.
Move to Episode 2, Framework, and decide whether you're running the loop yourself on the raw GenAI SDK or handing that job to Google's ADK.
Wire in Episode 3's function calling and Episode 4's browser control, giving your agent actual hands to click and navigate with while you're still talking.
Layer in Episode 5's memory (structured facts plus vector recall) and Episode 6's vision (snapshot versus continuous frame stream) without breaking the live feel.
💡 Pro Tip: Don't skip ahead to the browser automation in Episode 4 before finishing Episodes 1 and 2. The tool-latency problem covered in Episode 3 only makes sense once you understand what "live" actually costs computationally; skip that step, and the browser control episode will feel like magic instead of engineering.
Grab the code for the series and the Gemini Live API docs to follow along.
Stop Paying for 10 Tools. One AI Does It All.
Most e-commerce sellers are running their store across 6 to 10 separate tools — and spending more time managing software than growing their business. StoreClaw replaces your entire stack with one autonomous AI engine that monitors competitors, optimizes listings, automates marketing, and tracks real profit across Shopify, Amazon, and beyond.
It doesn't wait for you to ask. It runs 24/7 in the background, so you wake up to a full dashboard instead of a list of things you forgot to check.
Connect your store, and StoreClaw gets to work — no prompts, no complex setup, no six-app stack.
Free to start. No credit card required.
Classroom AI
Schools Are Teaching Kids to Catch AI's Mistakes, and the Results Are Already Ugly
Back to school this year comes with a new subject nobody asked for: teaching kids why the AI in their pocket can be confidently, spectacularly wrong.
A group of Charleston County, South Carolina educators got a live demonstration of exactly that. An instructor prompted an AI tool to "create a map of the world," and the result drew gasps of disbelief and laughter. Mali came back spelled "Mail." Egypt was labeled "Sopth." Libya was simply relabeled "Africa," with dozens of other countries misspelled or turned into gibberish, according to the AP report on the symposium.
Here's what's actually happening across the country:
Thirty-seven states have now published official AI guidance for schools, but South Carolina is not one of them, so the 50,000-student Charleston County School District set out to build its own approach from scratch, calling it "a huge endeavor."
The district brought in AI for Education, the same group that helped write guidance for Chicago, Houston, and dozens of other districts.
Utah's training model, teaching the basics of how AI works alongside bias, hallucinations, ethics, and data privacy, has already been copied by Maine, West Virginia, and Georgia.
Extensive AI literacy training requires steep investment, which raises real concerns that kids in under-resourced districts won't get the same attention as kids in wealthier ones.
One expert put the goal simply: good AI literacy includes knowing when not to use it.
Closer to home, FOX 5 Atlanta checked in on the metro Atlanta angle: Dave Howland, Executive Director of Teach for America Metro Atlanta, joined to talk through how local educators are already weaving AI into lesson planning and what families should keep in mind as the year kicks off.
Here's the uncomfortable part: nobody's really debating whether AI belongs in schools anymore. That ship sailed.
The real fight is quieter and uglier. It's about which districts can afford to do this properly and which ones are left improvising. This will sound familiar to teachers and administrators. A district with 50,000 students and a partnership with a national nonprofit can build a real curriculum. A district with a fraction of that budget gets a Google Doc and good intentions.
The literacy gap forming here isn't about AI at all. It's the same old resourcing gap wearing a new hat.
🔮 Prediction: Within two years, AI literacy becomes a required standard right alongside media literacy in most state curricula. But the states and private nonprofits moving first right now, not the Department of Education, will end up writing the de facto national playbook that everyone else copies. Whoever builds the first good template wins the standard by default.
Is your kid's school actually teaching this yet, or are you the one explaining hallucinations at the dinner table? Hit reply and tell me what you're seeing on the ground.
Vatican AI Policy
The Pope Just Told Lawmakers Exactly Where AI Crosses the Line
While Silicon Valley argues about scaling laws, Pope Leo XIV spent Friday telling an audience of actual legislators where he thinks the guardrails need to go.
Addressing the International Catholic Legislators Network at the Vatican, the Pope urged legislation that protects human rights and human dignity, warning that artificial intelligence must never be allowed to erode the family, which he called the "first school of humanity." Full remarks are covered by Vatican News.
What actually stood out:
He called for robust legal frameworks, independent oversight, informed users, and a political system that doesn't abdicate its responsibility, operating at local, national, and international levels at once.
He warned that AI's rapid development risks creating new forms of technological dependence, with poorer nations increasingly reliant on wealthier ones, and cautioned against innovation becoming "another vehicle for ideological or economic colonialism."
He named the harms directly: algorithms deciding who is seen and who remains invisible, platforms shaping public discourse without accountability, and worker dignity subordinated to system optimization.
He said AI must never erode the family, whether by reducing people and relationships to data and simulations, flooding young minds with content that distorts desire, or economic models that make family life precarious.
Here's the asterisk on this one: this wasn't a homily about screen time. It was delivered directly to people who write laws, at the exact moment governments everywhere are drafting AI legislation with no shared vocabulary for what "harm" even means.
The Vatican doesn't have a lobbying budget or a seat on any AI safety board, but it just handed legislators a ready-made moral framework: human dignity, family, accountability, at the moment they need one most.
🔮 Prediction: Expect language from this address to show up almost verbatim in upcoming EU and Latin American AI policy debates. Secular frameworks keep struggling to agree on shared values for AI regulation, so "human dignity" language from institutions like this becomes the fallback vocabulary lawmakers reach for when their own drafting stalls.
Should religious institutions have a real seat at the AI policy table, or is this overreach into a technical debate that isn't theirs to referee? Genuinely curious where you land; reply and let me know.
Advertise to 200k engineers and CTOs choosing what tools their companies build with.
Other Relevant AI News!
🧠 Anthropic's earliest institutional backer says the foundation model layer, not any single flashy feature, is still what's driving the entire AI wave. Watch Matt Murphy's take on CNBC as Anthropic heads toward a public listing.
🚀 Google DeepMind just partnered with the studio behind EVE Online to build agents that have to remember, plan, and scheme across years inside a persistent universe, no resets allowed.
📚 More than a dozen advocacy groups just asked the FTC to investigate AI companies for physically destroying books after scanning them, and they're not framing it as a copyright fight.
Golden Nuggets
🎒 AI literacy is becoming a real school subject, state by state, and the districts with money are the ones building it properly first.
⛪ The Pope just handed AI legislators a moral vocabulary they didn't have, built around dignity, family, and accountability.
🎙️ Google's omni-app roadmap shows what happens when agents stop taking turns and start perceiving, reasoning, and acting in one live loop.
Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)
Until our next AI rendezvous,
Anthony | Founder of Uncover AI