Stop learning to code. Learn plumbing instead.

Meta’s selling your AI chats, the $2.8T spending spree, and why this week marked AI’s shift from “amazing” to “not good”

Oct 03, 2025

Dear Friend,

While the world continues to do its thing, the news cycle this week surely didn’t stop. And as is the norm these last couple of years, AI dominates the headlines. What I find fascinating is that we have firmly moved on from the “oh my, look at what amazing thing AI just did”-phase to the “jeez, this is not good”-part of the party. As with every new technology, it takes quite a while until we truly calibrate where a technology ends up on the spectrum between hype and reality, or good versus evil. Point in case: OpenAI’s launch of Sora, their new video-generating AI. Only about a year ago, I am pretty sure the overall reception would have been one of awe and amazement. Today’s headlines are much more nuanced, with many concerned voices pointing out the potential for abuse (deepfakes) and the flooding of our feeds with AI-generated slop. Prepare for the many, many videos of shrimp Jesus singing Monty Python’s classic “Always Look on the Bright Side of Life.”

And now, this…

Headlines from the Future

Want a Long-Lasting Career? Become a Plumber or Carpenter! When I was still in Silicon Valley, we often joked (somewhat) about the fact that a) getting a plumber is nearly impossible, b) plumbers earn more per hour than the average developer, and c) plumbers might very well be the last profession standing once the AI and robotics tsunami has transformed all jobs. And, of course, the same is true for many other blue-collar jobs. It seems people are catching on to this:

Some 77% of Gen Zers say it’s important that their future job is hard to automate, with many pointing to professions like carpenter, plumber, and electrician as occupations they believe are safe from automation.

↗ As AI threatens white-collar work, more young Americans choose blue-collar careers

━━━━━

E-Commerce – AI is Coming For You! You heard it on our podcast first – in my interview with Scot Wingo, my former boss at ChannelAdvisor and now founder of agentic shopping startup ReFiBuy. Agentic shopping is coming for you, e-commerce! First, we saw product searches moving from Google to AI-powered search, such as Perplexity. Now, we are integrating the whole shopping experience into AI. This makes perfect sense and is, in many ways, a much better experience than traditional search.

↗ Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol

━━━━━

AI Movie Stars – The Backlash. First, we had AI influencers (hello, Lil Miquela), then music stars (welcome, Xania Monet), and now we have AI actors and actresses. The industry (the flesh-and-blood part of the industry, to be precise) isn’t too happy:

In a comment on the story on Instagram, Matilda star Mara Wilson asked: “And what about the hundreds of living young women whose faces were composited together to make her? You couldn’t hire any of them?”

↗ Stars speak out after AI actress attracts agency interest: ‘What about living young women?’

━━━━━

AI Regulation. One to Keep an Eye On. California (finally) did it. Governor Newsom signed a “landmark” bill requiring big AI companies to be transparent about their safety protocols. Keep an eye on this; we have seen similar legislation in Europe and some other places around the globe.

↗ California Governor Newsom signs landmark AI safety bill SB 53

━━━━━

Consulting Might Not Be the Glory Job It Used to Be Anymore. Accenture has started laying off employees “where reskilling is not a viable path for the skills it needs.” This makes one wonder how long it will be until Accenture’s customers realize that a significant portion of Accenture’s work can be accomplished by ChatGPT, Claude, Gemini, or any of the other numerous frontier models – for $20 per month?

↗ If you can’t use AI then it’s bye bye, Accenture tells staff

━━━━━

No, Quantum Computing Still Isn’t Real. HSBC (the bank) published a paper claiming “a 34% advantage in predictions of financial trading data.” Sounds too good to be true? Yes.

As they say, there are more red flags here than in a People’s Liberation Army parade. To critique this paper is not quite “shooting fish in a barrel,” because the fish are already dead before we’ve reached the end of the abstract.

Here is a good reminder that anyone who tells you they know how the future will play out has something to sell to you.

The paper can, more or less, openly admit all this right in the abstract, and yet it will still predictably generate lots of credulous articles in the business and financial news about HSBC using quantum computers to improve bond trading!—which, one assumes, was the point of the exercise from the beginning.

↗ HSBC unleashes yet another “qombie”: a zombie claim of quantum advantage that isn’t

━━━━━

Also, Dexterous Robots Aren’t Real Either. Every time Rodney Brooks writes, we listen. His latest takedown on why humanoid, highly dexterous robots are a must-read for anyone interested in the space.

In my opinion, believing that [humanoid robots will replace human workers] will happen any time within decades is pure fantasy thinking.

↗ Why Today’s Humanoids Won’t Learn Dexterity

━━━━━

Data on AI Models. The folks at Epoch AI published a huge data dump and analysis of over 3,000 AI models and the key factors driving machine learning progress. Fascinating!

↗ Data on AI

━━━━━

The AI Spending Spree Continues. Citigroup just raised its expected AI spending on infrastructure (think data centers and servers) to an eye-watering $2.8 trillion until 2029. That’s just a little over three years, meaning they expect spending to be around $1 trillion per year. I guess that’s great news for data center builders and Nvidia.

↗ Big Tech’s AI spending to cross $2.8-trillion by 2029, Citigroup forecasts

━━━━━

“We’ll see how this goes…” If you haven’t played around with OpenAI’s new video model and generator app, Sora 2, you ought to. The commentary from The Verge summarizes it well.

↗ OpenAI made a TikTok for deepfakes, and it’s getting hard to tell what’s real

━━━━━

And So the Inevitable Begins. Of course, it was only a question of time before AI followed Cory Doctorow’s “enshittification” curve. And of course, it had to be Meta, which leads the charge.

Meta is about to make your chats with its AI assistant part of its advertising machine, the company announced Wednesday. Beginning December 16, conversations with Meta AI — the company’s chatbot embedded across Facebook, Instagram, Messenger, and even its new Ray-Ban Display smart glasses — will be used to determine which ads and recommendations show up in your feed.

↗ Meta is exploiting the ‘illusion of privacy’ to sell you ads based on chatbot conversations, top AI ethics expert says—and you can’t opt out

What We Are Reading

🌱 This Company Is Turning Empty Offices Across America Into Indoor Farms Empty offices are becoming urban farms, bringing fresh food back to city neighborhoods. @Jane

📈 The Surprisingly Lucrative Business of Making a List of 500 Stocks Do you wonder how the S&P 500 works? This article explains its origins and how it came to serve as the main benchmark for the U.S. stock market and most retirement investments. @Mafe

💩 The Age of Enshittification It’s not just the apps, friends: We’re in an age of broad and creeping “enshittification” driven by deeply rooted dynamics and incentives in the economy, culture, and politics. @Jeffrey

🚀 Everything You Wanted to Know About ChatGPT Checkout and ACP Protocol Really interesting breakdown. It feels like ACP is doing for agentic commerce what HTTPS did for the web. This is the real starting line for agent-driven commerce. @Kacee

🚂 How Ruby Went Off the Rails As a former employee of one of the largest open-source companies, Mozilla, I find this a fascinating insight into the challenges of community-driven development and infrastructure. There are valuable lessons to be learned that extend well beyond the open-source community. @Pascal