The skinny on what's happening in AI - straight from the previous live session:

Highlights

🎨 Nano Banana 2: Finally Passes Wine Glass & Clock Tests

Discussed at [00:00:34]

Google's Nano Banana 2 in limited release, creating worryingly accurate deepfakes (Epstein/Trump/Diddy party images) and finally generates wine glass filled to brim and clocks at correct times!

Kyle's take: These have been the benchmark tests for years - no model could generate a wine glass filled to the brim (always half-full) or clocks showing anything but 10:10.

Nano Banana 2 smashed both tests - someone generated 11:15 on a clock AND a full wine glass in the same image. Whether they hard-coded these specific tests, we'll see.

The deepfakes are concerning as always. But the current model is pre-release and will certainly be guard railed and neutered by the time it launches.

🏛️ AI Breaks Government: £45 Planning Objections Flooding System

Discussed at [00:03:34]

Using an example called Objector (which charges £45 to generate planning objections) Ethan Mollick warns our current systems assume quality writing is costly and meaningful - not anymore!

Kyle's take: Our governmental systems - planning, courts, councils etc. etc. - assume writing and filing paperwork takes effort, so volume stays manageable. Well, more or less manageable! They are under strain already.

But now for £45, AI generates professional objections to your neighbour's extension. I built something similar on livestream for parking tickets - upload photo, AI contests it. Just a fun side project to give it a go. Took me maybe an hour to build?

We can now flood ANY system with AI-generated applications, submissions, complaints, rebuttals, even court cases. Every system that is built on the "writing = effort" assumption is about to break.

On our side (as people ahead of the curve) there are SO many businesses to build with AI. A great time for entrepreneurs!

Source: Ethan Mollick tweet

📊 McKinsey Report: 40% See ROI But 67% Stuck in Pilot Mode

Discussed at [00:09:07]

New McKinsey report: 90% of companies use AI, but 67% stuck in pilots. 39% seeing real EBIT gains. Top performers rebuild workflows, not just speed them up.

Kyle's take: 

I disagree with Greg a bit here. The 40% seeing EBIT gains is actually impressive for new tech - usually there's a learning phase where money gets wasted.

What’s most telling from the report is that the winners aren't just speeding up existing work - they're transforming HOW work gets done. They are baking AI

Google gets this: they're an advertising company (70-80% revenue from ads), not tech. They know generative AI kills search advertising, so they're cannibalizing themselves rather than letting competitors do it. That takes bravery. Most organizations pretend it's not happening and will become irrelevant. If you're watching this stream, reading my newsletter, you're in the 0.5% actually paying attention. We're in a bubble of people who care. Most people just use ChatGPT occasionally. We're still SO early.

Source: McKinsey AI adoption report / Greg Isenberg analysis

🎮 Why Do People Use AI to Cheat at Fun?

Discussed at [00:16:16]

Escape room players using ChatGPT to solve puzzles they paid $100+ to experience. AI cheating spreading to leisure activities?

Kyle's take: Last night at my board game meetup someone suggested using ChatGPT for CodeNames (lateral thinking game). A joke for sure but some people are legitimately using ChatGPT to “solve” fun activities.

People are using AI for pub quizzes, crosswords and more. An escape room worker watched twentysomethings pay over $100 then immediately ask ChatGPT for answers. As one TikToker said: "It's like going to a corn maze and wanting a straight line to the exit."

Have you found yourself doing this? “Cheating” on recreational activities with AI?

Source: The Cut article "Killjoy in the Machine"

Member Questions:

"Is there an AI for finding scholarships?" Asked at [00:06:08]

Kyle's response: Not specifically that I know of, but you could build one! I just checked and findscholarship.ai is available. Give people ability to input what scholarships they're looking for, their eligibility with a backend that has extended thinking model that searches according to needs. Then build database to cache queries for quick serving.

Sounds like a solid business opportunity (or just valuable free resource for students!) - so many AI businesses to build!! My problem now is focus - too many ideas, not enough time!

Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.

Streaming on YouTube (with screen share) and TikTok (follow and turn on notifications for Live Notification).

Audio Podcast on iTunes and Spotify.