Kia ora {{ First name | friend }},
We’re exploring the light and shadow of technology, with a focus on efficient AI workflows to free up your time.
This week’s edition features:
Google running circles around OpenAI
Nano Banana Pro image generation possibilities
Gene edited babies (?)
Group Chats in ChatGPT
AI’s Infinite money glitch
Thanks for being here.
Google’s Ecosystem Advantage Will Could OpenAI
A leaked internal memo from OpenAI CEO Sam Altman warns of “rough vibes” and “temporary economic headwinds” following the release of Gemini 3 Pro by Google. The memo admits Google’s recent AI work has shaken the competitive balance inside AI.
Google is taking things way beyond the chatbot game. It controls chips (TPUs), infrastructure, data from Search, Maps, Gmail, YouTube, and distribution via Android and Workspace. That gives it a structural edge over model-only players. Gemini 3 Pro reportedly surpasses OpenAI’s GPT‑5.1 capabilities on multiple benchmarks.
To me, Sam Altman’s memo is even more interesting because it drops his usual optimism. Instead he warns staff that OpenAI may face revenue slowdown, and suggests a shift toward “wartime footing”, focusing on superintelligence and infrastructure over short-term dominance.
What this suggests for the AI race: raw model performance no longer suffices. The winners will be those who own the full stack — silicon, data, infrastructure, and ecosystem. Google has built that stack. For others, the challenge is massive.
Expecting to see more of the “Everything economy” for companies to keep up with Google. Chatbots could feel like a cute nostalgia in the near future.
Featured Tool: Nano Banana Pro
AI Image generation just went to another level. I think this is what people have been seeking out when they’ve been thinking of AI images. The most common things I’ve heard are
It looks fake/generated
Brand/products are inconsistent
It can’t keep character consistency
It can’t do text well
Nano Banana seems to nail all of these. Far more accurate than midjourney, ChatGPT images or anything else I’ve seen. They’ve set a new benchmark which is also worrying, as it relates to image deepfakes, safety etc.
Here’s a few images I generated to test the above.
Location specific and realistic phone-styled images
Prompt: A selfie of rowan atkinson as mr bean, looking disappointed, drinking a bottle of tui beer in a fancy cafe in auckland nz
Product accurate images with real people in real environments
Prompt: A selfie of Riche Mounga drinking water from the classic blue bottle water from Anew in New Zealand as he sits on the sideline at Eden Park all blacks game. (This is using the bottle from my friend Jayden’s company Anew).
Generate receipts and KYC (Know Your Customer/Identity Verification) Images out of nothing. It even gets the math completely right.
Prompt: a receipt for a pound of cannabis from Snoop Dogg's store in Long Beach Los Angeles CA, for $420 usd excluding tax. Include line items for the cannabis, hemp papers and a 1:1 rolling tutorial with Snoop dogg. Total cost is $4,200 including tax
Nano Banana Pro Tips, Direct from Google:
Write prompts like a creative brief: clear subject, scene, mood, and purpose.
Describe composition and camera angle (close-up, wide shot, top-down) for control.
Use reference images to lock in consistent characters, products, or branding.
Iterate with specific edits instead of regenerating from scratch.
Be explicit with text: exact wording, font style, placement, and colour.
Keep prompts short but complete — 1–3 tight sentences work best.
Here’s a few interesting technology updates:
Nucleus Genomics is promoting gene-edited “designer” babies, with their ad campaign “Have Your Best Baby”. They have introduced new software that lets parents compare more than twenty embryos through hundreds of genetic indicators before choosing which one to implant. This allows new parents to select preferences in their babes ranging from height ranging from height and cognitive ability to mental health patterns and physical characteristics. 🤯 Check out Nucleus Genomics here.
GeoSpy AI can find location coordinates from an image anywhere in the world, even if it it’s from a reflection.. wild. Basically take a photo of a building, landmark or whatever, and it’ll find where it is. This isn’t using location metadata either, it’s algorithmic search using Google Maps 3D imaging. It can use low-context aspects of an image to find an exact building and orient the image to where it is. See the GIF below, and check their website here.
Group Chats are now available in ChatGPT, with the capability to add multiple people from the same team/project into a rolling context window. It follows the flow of the conversation and decides when to respond and when to stay quiet based on the context of the group conversation. Read more here.
AI Security Cameras are tracking staff efficiency in cafes, using motion tracking and analytical info to understand how quickly staff work, avoid breaks, and even understand exactly how long customers stay in store. Cool for business, weird for personal security and sovereignty if you expand this out to different areas of life… See more.
A founder of OpenAI says “School AI-detetction tools are doomed to fail”, outlining the impossible task of trying to monitor kids completing technology tasks/assignments with AI. Read more.
Here’s an example of how GeoSpy works
The AI Industry’s Infinite Money Glitch
The big boy AI giants are passing the same dollars around in a loop. Nvidia, AMD, OpenAI, and Oracle created a cycle where money never seems to leave the system. People are comparing this to 2008’s dot-com bubble, multiplied with insane leverage, and it raises questions about whether or not AI is actually a legit gold rush, or a false market boom caused by hype and stocks pumping.
Basically, the question is, does AI have any intrinsic value, or are we just being lead to believe it does? When company valuations are way beyond company earnings, the latter could be true.
Here’s how it works. Nvidia plans to invest up to 100 billion dollars into OpenAI. OpenAI then spends huge sums on millions of Nvidia GPUs. AMD enters with a 6 billion dollar gigawatt deal for their chips, and OpenAI takes warrants for a slice of AMD stock. Oracle buys Nvidia hardware to build data centers and sells the compute back to OpenAI. The same capital keeps circulating in a closed circuit.
Like these jokers:
The loop looks like an infinite money glitch from a video game. No clear profits. No clear end. Only bigger deals, bigger raises, and bigger bets.
Why this matters:
The flywheel boosts company valuations without delivering real returns.
Companies build massive infrastructure on hype-driven demand forecasts.
Investors chase exposure in fear of missing out on the next AI monopoly.
Market risk grows as the tech sector becomes dependent on circular spending instead of revenue.
Analysts say that this bubble popping could cause intense economic damage, amplified by global liquidity and AI urgency.
People are getting a bit shifty and warning that the bubble is expanding fast. The chip supply race, the capital recycling, and the velocity of deal making inflate expectations beyond what the market can support. The risk is not limited to startups. A sharp correction would ripple through cloud providers, chipmakers, and the broader tech sector.
I see this as a reflection of the capitalist western mind of “borrowing from the future” to serve now - sort of like going on a bender, then paying the consequences for a following week. Except this time we’re on a multi-year, multi trillion dollar AI hype bender, and the hangover could be decades of economic despair. I’m not qualified to make a comment on whether this is likely or not, but it seems that old mates at the top are saying “let’s get rich now, and let the chips fall where they may”.
{{ First name | friend }}, thanks for dropping in again.
Forward this to a friend if you found it useful.
Stay human,
Billy