On Tuesday, Spotify’s AI DJ learned to curse in four new languages. In Montreal, a Y Combinator startup called GovernGPT posted a job ad for engineers who want to build “thinking systems.” And at the Milken Global Conference in Beverly Hills, five people who collectively own the AI supply chain sat on a panel and admitted, more or less, that the whole thing might be built on a lie.
These are not separate stories. They are the same story.
The AI industry is simultaneously expanding its reach into your earbuds, hiring people to build machines that might replace the people doing the hiring, and publicly speculating that the underlying technology is wrong. Meanwhile, a privacy group called Noyb is suing LinkedIn for refusing to let users see who viewed their profiles. And Paul Krugman published a piece about insider trading in oil futures that has nothing to do with AI—except that it reminds us that when the rules are unclear, the people closest to the information always win.
Let’s start with the DJ.
Spotify’s AI DJ launched last year as a novelty: a synthesized voice that introduces songs with vaguely plausible patter about genres and moods. It was fine. Now it speaks Portuguese, Italian, German, and French. The company says the voice model was trained on a single real DJ’s speech patterns. The result is a machine that can tell you, in your language, why you should listen to a Bossa Nova track at 3 PM on a Tuesday.
This is not a big deal in isolation. It is a big deal when you consider that the same week, a startup is hiring engineers to build “thinking systems” that don’t just recommend music but make decisions about governance, compliance, and risk. GovernGPT’s job posting is vague on purpose—it promises “novel architectures for reasoning” and “systems that think before they speak.” The company is based in Montreal, a city that has become a weirdly important node in the AI world, partly because of academic talent and partly because rent is cheaper than San Francisco.
The Milken panel was more honest than most. Five people who touch every layer of the AI stack—chips, data centers, models, applications, regulation—told TechCrunch that the whole architecture might be wrong. Not “needs optimization.” Wrong. They talked about chip shortages, the absurd energy demands of training, and the possibility that the current approach to AI (massive models trained on everything) is a dead end.
One panelist reportedly said that orbital data centers—satellites with GPUs—might be necessary because there isn’t enough clean energy on Earth to run the next generation of models. Another said the industry is “burning cash to build something that might not work.”
This is the kind of thing that would be terrifying if anyone outside the industry were paying attention. But nobody is. The Spotify DJ keeps talking. The LinkedIn data keeps flowing.
Which brings us to Noyb.
The privacy group, founded by Max Schrems, filed a complaint against LinkedIn this week. The issue is simple: LinkedIn shows you who viewed your profile only if you pay for Premium. Noyb says that’s illegal under GDPR. The data belongs to you, not to LinkedIn. The company is essentially holding your social graph hostage.
LinkedIn will probably lose this case. But the damage is already done. The platform has trained millions of professionals to accept that their data is a commodity to be traded for access. The same logic underpins the entire AI industry: your music preferences, your job history, your search queries—all scraped, all fed into models, all used to build products that you then pay for.
Krugman’s piece on oil futures insider trading is a useful analogy. He argues that the financial system has become so complex and opaque that insiders can extract value with impunity. The rules exist, but enforcement is laughable. The same is true for AI data. The rules exist. Nobody enforces them.
So what do you do?
You could stop using Spotify. You could delete LinkedIn. You could read Krugman’s newsletter and feel righteous anger. But the machine doesn’t care. It will keep training on whatever it can find.
Or you could use a tool like Reportify AI, which cuts through the noise by summarizing the week’s most important tech news in a few minutes. It doesn’t replace thinking. It saves you time so you can think about the things that matter—like whether you want a machine to speak to you in your own language, or whether you want to know who’s been looking at your profile, or whether the entire AI economy is a house of cards built on stolen data and cheap energy.
The DJ is learning to say “thank you” in Portuguese. The engineers in Montreal are building something that might think. The panelists in Beverly Hills are wondering if any of it works. And your data is already gone.
Enjoy the music.