Trust Me, It’s $5 Trillion

A website manufactures fake revenue for laughs, publishers finally walk away from Google, and the AI you use so carefully is quietly rewiring how you think.

Jul 17, 2026

Dear Friend,

Andrew Ng, one of the leading (actual) experts in AI, just made a good point in the whole token-counting debate: AI tokens are cheap compared to human tokens (i.e., AI output is so much cheaper than your thinking time). Once you adopt this line of thinking, you start treating your (i.e., human) tokens as gold and AI tokens as a commodity. Read the thing – it’s good.

P.S. Here is a fun one – the wonderful folks at EDC recently had me on their podcast; it was such a fun conversation! Watch our conversation on YouTube or your favorite podcast app.

And now, this…

Headlines from the Future

Oh, the Absurdity of It All. I do like my LLM. Actually, I love it, and it’s insanely useful to me and my work. I also believe that there is a lot of potential in the further development and deployment of AI. But SoftBank’s Masayoshi Son going on the record to say that the AI buildout will require $5 trillion per year by 2040 and that AI revenue will hit 20% of global GDP by 2040 is just a bit too much for me…

“Every year $5 trillion, or 800 trillion yen, you might think that’s a lie, but I am confident that’s what it will cost,” Son said ⁠at SoftBank’s annual corporate conference in Tokyo. […] “The business model will be viable because by 2040, if AI revenue makes up 20% of global GDP, spending 800 trillion yen a year is a rounding error,” Son said.

The important bit here is this:

He did not say how he came up with the $5 trillion number or the proportion of global GDP he expects AI will make up.

Yep. Precisely. And just to put this all in perspective: Global GDP today is ~$117 trillion. 20% of that is ~$23T – larger than China’s entire economy ($19.4T) and about three-quarters the size of the US ($30.62T). All of worldwide manufacturing is about 15% of global GDP (and the single largest sector today).

━━━━━

AI Is an Environment, Not a Tool. Here is something to mull over: L. M. Sacasas makes the argument that AI reshapes your perception and judgment from the inside, even when you use it carefully or reject its every suggestion. This leaves you in a position where even carefulness offers no protection – the most deliberate users lose judgment fastest because the medium, not the content, does the reshaping.

But AI is not a tool in this sense, it is an environment which envelops the user and works on us from the inside out while we naively think that we remain unchanged by our use so long as we are using it carefully and intentionally. The care and intentionality is beside the point, and our confidence in such vigilance probably works against us in the long run. […] And this is because AI is an environment not a tool. I can pick up a tool and put it down, but the environment absorbs me into itself.

What We Are Reading

The End of Reading Is Here A world where everyone could read once felt like progress locked in. It’s starting to look more like a temporary experiment. @Jane

Scientists Say You Have a Sixth Sense, and It Could Affect Your Mental Health Interoception is a hidden sense we all have that helps regulate hunger, mood, and anxiety. Some of us have it more developed/trained than others. @Mafe

Terrified of Small Bets Your teams (and your org) aren’t locked in by a lack of imagination or tools. More likely, they’re trapped by systems that require “certainty” before action. @Jeffrey

Transforming Manufacturing at Pfizer: The Hard Part Was Not the Technology One of the better AI case studies I’ve read lately. Pfizer shows that getting the technology working is often the easy part; changing how thousands of people work every day is where the real transformation happens. @Kacee

The Most Famous AI Writing Tic Is Also the Most Mysterious I hate the AI trope of “it’s not X, it’s Y” – thank Shakespeare for that… but really, nobody knows why LLMs love this construct so much. @Pascal

Down the Rabbit Hole

🧑‍⚕️ The tl;dr of this new study: GPT-4o, given the right prompting strategy and note-selection freedom, generates ED “one-liner” clinical summaries that emergency physicians rate as more accurate, complete, and clinically useful than the physicians’ own summaries (and this is with GPT-4o, a model so old that it’s (largely) not even available anymore).

📚 The tides are turning: With Google Search traffic collapsing, publishers are now starting to withold content from Google, as the trade-off equation has flipped: Once unimaginable, publishers are preparing to opt out of Google Search.

💵 It surely is not too surprising that OpenAI will argue you shouldn’t count tokens, but focus on work per dollar… How to manage AI investments in the agentic era.

👷 Poof goes the data center. New York’s ban on data center buildouts has the potential for far-reaching consequences: New York imposes first-in-the-nation statewide freeze on ‘hyperscale’ data centers.

🤨 Is anyone who has spend any time on LinkedIn or X even surprised anymore? LinkedIn and X are flooded with AI spam, browsing data suggests.

😙 Talking about LinkedIn – its seemingly inevitable decline into Facebook’s ugly stepbrother continues: LinkedIn, a mass grave of ghost jobs, is now becoming a dating app.

😯 And taking it a step further, social media seems to be dead now anyway (at least the “participating” part of it): The death of the status update: Why 55% of Americans stopped posting on social media.

💈 It’s not you, it’s the carpet: Striped floors and flickering leds can overload the human mind, leaving some with headaches or nausea.

🍞 The bread paradox: why convenience always wins, and why SaaS isn’t doomed.

🧑‍🏫 The U.S. spent $30 billion to ditch textbooks for laptops and tablets: The result is the first generation less cognitively capable than their parents.

🔋 The energy transition is full steam ahead: In three states in Australia you can now get three hours of free daytime electricity.

😱 Fascinating thought experiment: What would happen if WhatsApp were to disappear? The day WhatsApp goes dark.

🚶 Those who wander are not lost. But we seem to loose the ability to wander (at least – and the consequences are real: The problem with Google’s A.I. overview.

👽 John Carpenter’s master piece “They Live” was right all along: Guerrilla london bus ads mock Kylie Jenner’s Meta glasses campaign.

🕶️ Even Lorde says so: Lorde says Ray-Ban Meta AI glasses are ‘not sexy.’

🤑 This is too funny: “LARP – Revenue is just an agreement between friends” is a website which allows you to generate some good old fake revenue by doing some nvidia-style circular deals…

🎬 Fun one for the weekend: The tech of ‘Terminator 2.’

↗ Dive into the deep end: Access our complete collection of 2,900+ radical links.

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.