👋 Hello hello,

Hope Monday is treating you well.

So here's something worth sitting with: 90% of parents believe their kid is at grade level. Only 28% actually are. And instead of closing that gap, kids are now using AI to skip the thinking process entirely.

Meanwhile, ChatGPT shipped something that everyone who's ever had a "one quick question" spiral into a two-hour rabbit hole will really appreciate. And Google dropped a creative agent inside Flow that changes the game for anyone making content at scale.

Also, Kushank flew to Toronto and spoke at Google's AI Gatherings during Tech Week. We'll get to that. Let's dig in.

🔥🔥🔥 Three Curated AI Updates

Brilliant launched an AI tutor called Koji, and it's taking direct aim at a genuinely alarming stat: American kids can't read or do math anymore. The product is trained by learning experts from MIT and Harvard and built specifically to coach kids through problems — pointing, sketching, and annotating in real time — instead of just handing them answers. Think of it less like a calculator and more like a patient tutor sitting next to your kid who refuses to do the homework for them.

The timing matters. AI cheating in schools isn't a future problem — it's already here. Koji's entire bet is that the right response is AI that makes kids think harder, not less. That's a harder product to build, and if it works, it's probably the most important AI education tool launched this year.

No confirmed rollout date yet, but the product is live and taking sign-ups now.

If you've ever started a ChatGPT thread with "one quick thing" and ended up 40 messages deep with no idea where you put that one useful response — this one's for you. OpenAI rolled out a table of contents feature for conversations with five or more exchanges. It auto-generates a navigable index so you can actually find what you need without scrolling back through everything.

It sounds like a small quality-of-life update, and it is. But it's also a sign that OpenAI is taking the long-conversation use case seriously. Power users who live inside ChatGPT all day will feel this immediately. Available now.

Google shipped a new agent inside Flow — its AI filmmaking and creative tool — and it's more capable than the name suggests. The Flow Agent can brainstorm with you, punch up dialogue, suggest plot directions when you're stuck, generate multiple scene variations in parallel, and then organize and rename all your final assets when you're done.

The whole thing runs on Gemini models and is designed to stay in service of your creative vision rather than taking it over. For solo creators and small production teams, this is the kind of end-to-end creative support that used to require a writers' room. Worth experimenting with if you're building any kind of video or narrative content.

🔥🔥 Two Pro AI Resources

Anthropic just published a free security framework specifically for teams deploying AI agents in the enterprise. Zero Trust, if you're not familiar, is a cybersecurity model built on one core idea: trust nothing, verify everything, assume a breach has already happened. This guide applies that thinking to agentic AI — so things like how to give agents minimum necessary permissions, how to detect when an agent is being manipulated or misused, and how to build deployments that can contain damage when something goes wrong. If you're a CISO, architect, or anyone responsible for AI infrastructure, this is required reading. It's free, practical, and covers threats that most teams aren't thinking about yet.

Kushank was invited to speak at Google's AI Gatherings event during Toronto Tech Week alongside Amit Vadi from Google DeepMind and Smitha Kolan from Google Cloud. The whole conversation is worth watching — they get into the difference between builders and developers, why you shouldn't jump straight into complex agentic frameworks, how skills work inside agentic systems, and where the real opportunity is right now for people who understand these tools deeply.

Great hour of substance if you want to understand where Google's AI stack is heading.

🔥 Learn How To Create a Drone Shot Video Using AI

If you have a photo of any location and you sketch a rough flight path on top of it, Gemini can turn that into a cinematic FPV drone shot. No drone. No footage. Just a still photo, a drawn line, and the right prompt.

Here's how to do it:

  1. Take or find a photo of any location — a skyline, a street, a landscape.

  2. Open it in any basic image editor (even your phone's markup tool) and draw a red arrow or curved line showing the path you want the "drone" to follow.

  3. Go to Gemini and upload the marked-up photo.

  4. Paste the prompt below exactly as written.

  5. Gemini will generate a cinematic FPV video following the path you sketched — erase the red lines, render the shot.

Here's the prompt to use 👇

Please erase the red lines, arrows, and all auxiliary markers. The red lines and arrows are only for reference as camera movement paths and must not appear in the final footage. The shot is presented in a first-person FPV perspective, ultra-high-speed, cinematic, one continuous take, strictly following the designated route. The shot starts from street level near the right skyscraper's base, passes [continue describing your path here based on your sketch]

Login or Subscribe to participate

Until next time,
Team @PracticalyAI

Reply

Avatar

or to participate

Recommended for you