Microsoft's free GEO tool / Hyperscaler debt drama
Good morning!
Yesterday I was at my favorite cafe and a guy sat across from me with a t-shirt that said, “Do not talk to me about AI, I will kill myself.” I laughed. Irony: he then started talking about AI with his family.
AI: It’s a weird era, and today’s infants will admire us for living through its roll-out. Maybe!
Have a great week!
Kos
Microsoft drops Topic Insights, a great free tool to see how a brand shows up in AI
This looks like a real win, a free tool that gives you a report card on your GEO. Also tells you how your competition is doing.
OpenAI is offering tutorials for small biz owners, including marketers
Sounds cool, but they didn’t say in the post if it’s free. This is what happens when you don’t hire copywriters.
AI is compressing preclinical drug-development timelines by 70%
A survey finds that new drug development programs could grow by 10% in 3 to 5 years. Yay for science.
The scroll-world agent skill turns landing pages into an immersive scrollable world
Works for Claude Code, Codex and other agents. Very cool fly-through effect, but Claude says it’s a bit of a data hog, but Claude also makes mistakes. There is a video on the GitHub page that shows what it looks like.
A report on the extent of corporate AI psychosis
That’s the weird thing about having a white collar job these days: the AI stuff feels crazy, and I say that as a huge fan of Claude Code. We are in a paradigm where people who point out that AI “solutions” are crap are penalized. Many people in the trenches can tell that the AI enthusiasm is not translating into sensible rollouts and execution, but … it’s bigger than that. Software engineers are doing the work they are paid to do, but then they are having AI re-do the same work because their bosses will scream at them for not using AI. Psychosis is an accurate term.
Claim: we will turn video into something your AI can search and understand
VideoDB claims that they can turn video clips and live streams into data that AI agents can search and understand. AI can search and retrieve exact moments in the video and turn the whole thing into persistent memory. Even if VideoDB’s product doesn’t deliver, someone will nail it soon.
Micro-influencers are blowing up and attracting more marketing dollars
Micro-influencers are becoming a new middle class. Marketers like micro-influencers’ high engagement rates. Also, they don’t need millions of followers. People are landing 4-figure deals with as few as 8,000 followers. Those with 100,000 followers can earn $180,000 per year. Bonsa Nemera has no financial credentials but people love his financial advice, and he has inked deals with Intuit, Cash App.
Five hyperscalers have $1.6 trillion in hidden debt
That’s not a typo. Trillion. Meta, Alphabet, Microsoft, Amazon, and Oracle are using special purpose vehicles and off-balance sheet arrangements to hide it. And this is on top of the $1.35 trillion of debt those companies reported for the most recent quarter. Meta alone is on the hook for $420 billion of the total. Fun fact: debt hidden in shell companies caused the collapse of Enron.
LinkedIn and X are packed with AI spam
Pangram has a chrome extension that flags AI slop (I haven’t tried it but I’m curious). The company says 1 in 4 longform posts on social media are fully AI-generated. Of all posts flagged, two-thirds were from LinkedIn. Gross.
Having one model check a different model's work for errors is more effective
Claude does a better job of finding code problems with Codex code than it does analyzing its own code, and vice versa. If you want to boost your odds of catching code problems, have a different model analyze your code.
Keys to successful roll-out of AI tools
First: ensure people are comfortable publicly being confused and making mistakes. Second: start with small, reversible experiments instead of committing to one massive project. Third: work through projects across departments.
Jack is launching Buzz, a mash-up of Slack and GitHub
It is trying to bring AI agents, code repositories, and team discussions under one roof.
AI is forcing UX designers to start designing again
Somehow over the years actually designing interfaces became second-tier work, subordinate to building strategy decks. AI is forcing designers to once again design, to prove they can do it better than Claude Code.
AI companies are buying (and destroying) tons of books to train models
AI companies have started buying books published prior to 2022 to train their models because those books are guaranteed to be free of AI slop. Scanning them for training destroys the books, some of which are rare. There are websites that allow AI companies to make these purchases in secret to avoid headlines like “Anthropic is destroying millions of books.” Paywalled.
BrainCo lets you control a robot with your brain
The company showcased the technology at the 2026 World Artificial Intelligence Conference. You put an EEG headset on, and then you can make robots do stuff.
Publicis is hiring while other holding companies are firing
They seem to be doing well by investing in people, they just hit their 21st consecutive quarter of growth, which I think refers to revenue growth. Omnicom and WPP are both laying folks off.
A shift from giving fans more content to giving them more ways to participate
Little Black Book →Today in schlossmaxing
This website is a big ol’ map of the world’s castles and fortresses.
You might have this alarm in your car; a hacker can wreak havoc with it
There is a type of alarm that car dealerships sometimes install to prevent theft from their lots. Frequently the alarm stays on the car after purchase without the buyer knowing it. Hackers can get into it and turn your car off and cause other mayhem. To see if you have the alarm in your car, look for a KARR or a SWDS sticker on your windshield.
