Dear Friend,
As I am writing this, the world is debating if the world is ending. In case you missed it – a former OpenAI and Anthropic employee went on the record to state that there is a 10% chance that AI will destroy humanity by the end of the decade. While emotions are running high, I encourage you to calibrate news like this with counterarguments, e.g., from AI researcher and LLM critic Gary Marcus (here and here).
And now, this…
Headlines from the Future
Stock Trading? Let My AI Handle It. First we had robo-advisors (and they somewhat upended the traditional financial advisor market); now we have vibe-coded trading AI agents.
Like many of Wall Street’s top money managers, Colin Edsman has a team helping him stay on top of his investments. There is Alex, who scans the market regularly for promising stocks and exchange-traded funds. Sarah, who checks over open positions each day before the close. Elena, who whips up a weekly performance report. There is just one difference: Alex, Sarah and Elena aren’t coffee-chugging analysts pulling all-nighters at their terminals. They are Claude agents, working round-the-clock from Edsman’s laptop on his kitchen table. And so far, they are outperforming most of the accounts he runs himself.
Sounds great, right? Until it doesn’t. The financial advice forums where people report on their use of AI-powered trading agents are full of horror stories of agents going off the rails, doubling down on bad trades and wiping out accounts. And yet – people like Edsman are pushing the boundaries.
━━━━━
The Slop Deluge: A recent investigation from Semafor found that 50 out of 310 op-eds published in the most prestigious outlets in the US (WSJ, NYT, and WaPo) were flagged as at least partially AI-written by a detection tool. An op-ed is the last place you want slop to show up. On the flip side is the first community-steered AI slop “TV station” – an endless stream of AI garbage that is strangely addictive.
China’s AI Job Market Challenge: US and European headlines are full of stories on how tight the job market is for recent college graduates (and, interestingly, the job market in the US for non-college graduates is the best it’s been in two decades!). Now imagine you are graduating in China – into a job market which is under the same pressure from AI, but there are a record-breaking 12.7 million of you. Youth unemployment in China is at 17.9 percent, which leaves leadership in Beijing with a real problem, because the party’s social contract has always been “jobs in exchange for staying quiet.”
Build a Data Center, Have to Insure It: Here’s a business opportunity for you – as much as the news is all about the great data center buildouts in the US (and the resistance from the communities where they are built), these data centers need to be insured. And all too many of them face a long list of threats both during construction and after – from extreme weather to power outages, all the way to terrorist attacks. And as data centers tend to be built in clusters, many of these risks are shared across multiple sites – potentially triggering large payouts (which insurance companies make their clients pay for in the form of high premiums).
Papal AI Governance: Leave it to the Pope to inspire a framework for ethical AI governance. Researchers at the John Paul II Catholic University of Lublin have developed a framework based on the Pope’s encyclical Magnifica Humanitas. The eight principles are recognizably Catholic (human dignity, solidarity, subsidiarity), with that last one (subsidiarity) being one you don’t see in other frameworks: Push every decision to the lowest level competent to make it. The frontier labs have no equivalent rule, and it shows.
The Compressed Cybersecurity Timeline: As increasingly powerful, yet easily accessible and open AI models ship, like the Chinese GLM 5.3-flash, it gets easier and easier to find and exploit vulnerabilities in software. The race between frontier models (some of them withheld from the public for security reasons, like Anthropic’s Mythos) to find and patch vulnerabilities and open, modifiable (read: one can remove the security safeguards) AI models is heating up dramatically as release timelines are compressing. System engineers like Jynn Nelson are concerned – for good reason, as we might be running out of time to patch the vulnerabilities in our software before the bad guys get a chance to exploit them.
Anthropic’s Economic Scenarios: The company which brought us Claude published its model of the US economy in 2030 – packaged in three scenarios (modest/substantial/extreme). Growth shows up in every scenario, but in the extreme one, GDP ends 32% higher while labor’s share falls from about 60% to 45% and knowledge-worker wages drop more than 10%. Economists were quick to point out flaws in the methodology – but it’s still illustrative, if for nothing else but to understand how one of the leading AI companies, which is shaping our future, sees it.
What We Are Reading
The Myth of Perfect Knowledge Certainty was always an illusion, and understanding that might be the most important thing science has taught us. @Jane
I Study Successful People. They Have 1 Habit in Common. We suffer from the illusion of self-sufficiency, overlooking how much we owe our skills to other people. And the most successful people understand that a willingness to ask for help is crucial to achievement. @Mafe
The Network Always Beats the Castle We all intuit that the AI revolution will change the shape of firms and flows within them, but we don’t yet know how. An analogy from an earlier tech revolution might show the way. @Jeffrey
AI Is Blurring the Line between Sales & Marketing I’ve always viewed Sales & Marketing as a marriage; two distinct disciplines that should operate as one, but too often live separate lives under the same roof. This HBR article explores how AI may finally force the union, transforming two loosely aligned functions into one connected growth engine. @Kacee
How American Fitness Became a Cult of Self-enhancement Our quest for longevity and overall health can turn a dark corner when our fitness tracker scores become more important than the quality time we spend with loved ones or the things we truly enjoy doing. Go ahead, have that drink with your friend at the bar and don’t worry too much about your sleep score. @Pascal
↳ And now for some fun…
⚛️ Explore all the world’s nuclear power plants in this interactive 3D map. Did you know that there are a grand total of 447 operating nuclear reactors in the world?
What We Do When We’re Not Writing
Hi! I’m Pascal from radical. When we’re not writing this newsletter, we help organizations turn volatility into advantage – without the “innovation theater.” We build your team’s capacity to handle disruption, so you stop reacting and start shaping. If that sounds useful, let’s talk.
↗ Dive into the deep end: Access our complete collection of 3,100+ radical links.
I cannot help but eyeroll at the whole AI trading thing. As if everyone forgot about credit default swaps and mortgage-backed securities and wants to make 2008 great again, when nobody understood what assets were toxic and what their money was actually invested in.