The AI Efficiency Trap

Amazon abandons the physical world, Europe declares war on Visa, and the UK’s disastrous approach to automated labor

Dear Friend,

While Jane and I were in a galaxy far, far away (pictures here in case you’re curious), the AI world went bonkers over ClawdBot/Moltbot/OpenClaw – the open-source “autonomous agent” that acts like a personal virtual assistant. Some hailed it as the first “true” AGI. It was (and continues to be) a security nightmare. It also doesn’t work. Or it works very well. Depends on who you ask. But as quickly as it came, it also disappeared (technically in less time than it took Jane and me to fly to Patagonia, sail and climb in the Darwin Range, and come back home). Another good reminder that nothing is eaten as hot as it is served (as the Germans say).

And now, this…

---

Headlines from the Future

AI Doesn’t Reduce Work – It Intensifies It. UC Berkeley researchers spent eight months studying 40 workers at a 200-person tech company to see what actually happens when you give knowledge workers access to AI tools. And what they found should dampen your excitement about the “AI-enhanced human”: Rather than reducing workloads, AI created a self-reinforcing cycle – it accelerated tasks, which raised speed expectations, which increased reliance on AI, which widened the scope of what workers attempted, which further expanded the quantity and density of work. In sum: Workers weren’t told to do more – they chose to, because AI made “doing more” feel possible and even rewarding. The result was faster pace, broader scope, and longer hours, all driven by the employees themselves.

It would seem that since AI increases productivity, it means you save time and work less. But in reality, you don’t work less. You work the same amount or even more.

Talk about hacking your internal reward system… and you thought social media was bad. Or, in other words: The treadmill just got faster.

P.S. This post by Steve Yegge is making the rounds – same idea. “The AI Vampire”

━━━━━

AI Is Cutting More Jobs in the UK Than Anywhere Else. A new Morgan Stanley report compared AI’s impact on employment across the US, UK, Japan, Germany, and Australia – and the UK stands out for all the wrong reasons. British firms reported an 8% net job loss linked to AI, double the international average. Meanwhile, UK companies saw roughly the same 11.5% productivity boost from AI as their peers in other countries. American firms with similar gains actually created more jobs than they cut (at least according to the report – and at this moment in time). Same technology, same productivity uplift, very different choices – which tells you this is a management story, not a technology story. To make it worse, UK employers were most likely to axe early-career positions requiring two to five years of experience, hollowing out exactly the layer where people build the skills they’ll need for the next three decades.

Executives are conflating early tool investment and adoption with license to reduce headcount, often before demonstrating genuine productivity gains. UK boardrooms appear particularly susceptible to cutting first and measuring later.

Not good. And also – “AI Washing” is a thing.

━━━━━

Amazon Gives Up on Amazon-Branded Grocery Stores. Amazon is shutting down all 72 of its Amazon Fresh and Amazon Go locations – the company’s decade-long attempt to crack physical retail under its own brand. The closures are effective February 1, which means by the time you read this, they’re likely already gone. Amazon’s pivot: double down on Whole Foods, which has grown 40% since the 2017 acquisition and is expanding to 100+ new locations, plus a new “supercenter” concept in suburban Chicago slated for 2027. It’s a quietly remarkable admission – the company that redefined how the world buys things online could never quite figure out how to make people walk into a store. Remember, this is a list that includes bookstores, 4-Star shops, electronics kiosks, and a clothing store called “Style” that lasted all of two years.

While we’ve seen encouraging signals in our Amazon-branded physical grocery stores, we haven’t yet created a truly distinctive customer experience with the right economic model needed for large-scale expansion.

Talk about corporate-speak for “it didn’t work.”

━━━━━

Europe’s Done With Visa and Mastercard. Europe is finally getting serious about (payment) sovereignty (undoubtedly driven by the overall political climate). The European Payments Initiative’s digital wallet Wero – built on SEPA instant credit transfers, no card required (it’s actually quite neat – all you need is a mobile phone number), no American intermediary – already has 47 million registered users across Belgium, France, and Germany, and is about to add another 130 million users across 13 countries. Running in parallel is the ECB’s digital euro project. The strategic logic is straightforward: when Visa and Mastercard cut Russia off in 2022, European policymakers realized that American payment infrastructure can be weaponized – and every transaction routed through it sends European consumer data to the United States. ECB President Lagarde has called the situation urgent. Mastercard’s CEO says he’s “not particularly worried.” One of them will be wrong.

European payment sovereignty is not a vision, but a reality in the making.

We’ll see. But the direction is clear.

━━━━━

The AI User Gap Is Astonishing. Martin Alderson makes a simple but important observation: two kinds of AI users are emerging, and the gap between them is enormous. The first group is all in – using Claude Code, MCPs, agentic workflows, the whole stack. Surprisingly, many of them aren’t technical at all; Alderson has seen finance people getting extraordinary value out of AI precisely because Excel is so limiting compared to a full programming environment like Python. The second group is chatting with ChatGPT occasionally and calling it a day. This split, Alderson argues, explains a lot of the confusing media coverage about whether AI actually boosts productivity – it does, dramatically, but only for the people who’ve crossed a usage threshold that most haven’t.

I am still shocked by how much difference there is between AI users.

This tracks with everything I see and hear. Good read. (and in related news: Your AI adoption metrics are lying to you.)

---

What We Are Reading

What Technology Takes From Us – and How to Take It Back Technology is increasingly becoming central and critical to our daily lives. Is it time to take our humanity back? @Jane

Inside Dollar Tree’s Push to Lure Rich Shoppers Hunting for Bargains It’s hard to say no to a good deal – shoppers who make over $100,000 are driving much of Dollar Tree’s current growth. Last quarter, 60% of new Dollar Tree customers made at least six figures. @Mafe

AI Is Getting Scary Good at Making Predictions AI superintelligence may (or may not!) still be a few years away from being a few years away, but AI superforecasting – a different but still highly valuable data- and modeling-driven proposition – seems to be very close at hand. @Jeffrey

Forbes 250: America’s Greatest Innovators Celebrating the minds shaping tomorrow, because true innovation isn’t just about invention, but rather impact. Spoiler alert – Elon beat Bezos. @Kacee

Why the Moltbook Frenzy Was Like Pokémon The “Moltbook” social network for AI agents, while hyped as a glimpse into the future of autonomous AI, was actually more akin to a chaotic game of “Twitch Plays Pokémon” – a spectator sport for humans rather than a functional hive mind. @Pascal

---

Down the Rabbit Hole

💼 Here are some useful patterns to follow when implementing AI agents in your workflow.

🏎️ Not saying you should do this, but it is pretty cool: Comma AI’s active driver assistance system (Comma AI is founded by famed hacker George Hotz, and is using an Open Source approach to their work).

🙊 The end of your voice as a unique identifier: Chinese AI model Qwen3 TTS needs just a few seconds of your voice to generate a convincing voice clone.

👨🏼‍⚕️ The headline hides the punchline: “I let ChatGPT analyze a decade of my Apple Watch data. Then I called my doctor.” – turns out, AI is an absolutely terrible doctor. Which shouldn’t come as a surprise. But just in case… And if you need more data – here’s a new study on the subject (same conclusion).

🌀 Now we finally know (with mathematical precision) when the Singularity will happen: Tuesday, July 18, 2034 at 02:52:52.170 UTC. Set your watches.

🗽 This is as insane as it is delightful: New York City as a massive isometric pixel landscape running in your browser.

🥾 The good, old Internet is still alive: A delightful almanac for all things laces – as in “shoelaces.” Everything and anything you ever wanted to know (and not) about laces.

↗ Dive into the deep end: Access our complete collection of 2,500+ radical links.

---

Pascal is getting back into the swing of things, after being in an environment completelty void of humans.

---

Should We Work Together?

Hi! I’m Pascal from radical. This newsletter is our labor of love. When we’re not writing, we run radical, a firm that helps organizations navigate the future without the “innovation theater.” Most leaders want to seize new opportunities, but they hate endless strategy decks that go nowhere. At radical, we don’t run “projects”; we build your organization’s internal capacity to handle disruption and change. Our goal is to make you future-proof so you can stop reacting to the world and start shaping it. If you’re interested, let’s jump on a call to see if we’re a good fit. Click here to speak with us.