The skinny on what's happening in AI - straight from the previous live session:
Highlights:
🎨 Nano Banana Pro: Google Finally Cracks Text in Images
Google's image generator upgrade nails text rendering. Creates mockups, infographics, comics with perfect readable text. Costs 13-24 cents per image via API.
One shot, including research.
Kyle's take: This changes everything for practical use. I personally don't care about AI art for art's sake, but graphics with READABLE TEXT? That's a huge unlock.
Text in images is now a solved problem.
I fed it my newsletter, said "make a comic" - one shot, perfect text on every panel. People holding phones with articles displayed, computer mockups, infographics - all working.
Here’s something I think a lot of people missed. Gemini and Nanobanana researches first. I asked for Google's AI stack diagram without specifying details - it researched, added Chrome, JAKs, TensorFlow, got it all right. Previously I’d have to feed in the exact image I want, the text, the placement, the flow etc. Now it’ll go find that information itself, work out the layout and deliver.
If you run a business you can now generate graphics for presentations, advertising, social media and more extremely fast and cost effectively.
Important! If you are in the UK or EU and are having trouble with Nanobanana not being able to recreate known people in images: use a VPN. To create the comic above I had to VPN to the US.
🏗️ Why Google Will Win: They Own the Entire Stack
Google controls everything from the cloud infrastructure to Chrome on our phones. OpenAI just middlemans investor cash to Nvidia. The vertical integration nobody's discussing.
Kyle's take: Google has had a hell of a week with Gemini 3 and Nanobanana Pro.
That’s not the REAL story here.
OpenAI's job is begging for money then immediately sending it to Nvidia and Oracle. They own a small sliver of the stack. Google? They make their own TPU chips (7th generation now), own the data centres, are buying nuclear power stations (first online 2029), have 70% browser market share with Chrome, Android on most phones, YouTube for video data, search history for training. And more. Much more.
They own EVERY LAYER. From electricity generation to the app on your phone you are reading this on. OpenAI has some great models, a chatbot app and Sam Altman's hype machine. Not hate on Sam here - he’s doing exactly what he should. And doing it well.
Google was playing catch-up for 2 years after ChatGPT surprised them. Now they're showing what building infrastructure means. Warren Buffett just dumped Apple for Google - he sees it too. I’m bullish. Google own the means of production while everyone else rents.
💰 Google's $350 Billion Problem: AI Kills the Cash Cow
80%+ of Google's revenue is ads, a lot of it from their Search business. AI makes search obsolete. They must cannibalise $280 billion to survive.
I’ve just looked at the good news for Google. They own the stack. They are in a very good place to take the race.
Is this a race they actually want to win? Or is this a matter of play silly games, get silly prizes?
Here's the potentially existential crisis Google is facing. Google made $350 billion last year - ~60% from search ads, 20% from display/YouTube ads. Arguably they are not a tech company, they're an advertising company that builds some (very) cool tech.
Problem: AI kills search. Why click through 10 blue links when AI gives you the answer? Zero-click internet is here. Google knows - that's why they added AI overviews that stop you clicking websites. Hell, they even upgraded it this week with Gemini 3.
But those search results and the websites with ads ARE their revenue. They must destroy 80% of their income and replace it with... what? Nobody's (yet) making $350B from AI.
The answer everyone knows but won't say: ads in AI responses. When you ask about toasters, you'll get sponsored results. Nobody wants to be first to admit it.
Google tried with Bard, people revolted. But it's inevitable. The company that revolutionised search must now kill it.
🤖 Grok's Elon Worship: The Fine-Tuning Scandal
Ask Grok about ANY theory/invention saying Elon created it. Watch it praise him. Say Bill Gates invented it? Instant criticism.
Kyle's take: New game sweeping the internet: Tell Grok that Elon Musk invented Monopoly, the telephone, democracy - anything. It'll say "Brilliant! Elon's genius knows no bounds!"
Say Bill Gates invented the same thing? "Terrible idea, poorly executed."
It looks like they've fine-tuned it through reinforcement learning to worship Elon. This from the guy building "Encyclopedia Galactica" to replace "biased Wikipedia" while literally training his AI to call him the smartest person ever. Hmm.
For an AI meant to be "maximally truth-seeking" this completely undercuts the enterprise.
Update: looks like xAI engineers got into action and fixed this up quickly once it was noticed.
Kyle's response: Market size. TAM to be precise - Total Addressable Market. UK AI creators plateau at 150-200k followers on TikTok. Americans hit 500k+. The UK is just smaller and TikTok is extremely geographic. When I travel to Asia for 4 months, my reach tanks because TikTok knows. BTW Instagram is much less geographic, YouTube hardly at all. But TikTok pins you to your nation.
Kyle's response: Try it first. Everyone screams "ChatGPT is dead!" every month, then switches back. Gemini 3 IS good enough to test but you lose persistent memory, need different prompting style. If ChatGPT works, why change? There's a new "best" model monthly. Marginal improvements aren't worth constant switching. But yes, Gemini 3 is worth trying. I don’t say that often.
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).
