When AI Eats Itself
While datacenters drink our water supply, Moore’s Law reverses course, and Meta quietly abandons the metaverse.
Dear Friend,
I am constantly reminded of the F. Scott Fitzgerald quote, “the test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time, and still retain the ability to function,” when I think about AI. Without a shred of doubt, AI is the most overhyped technology of our time. Most of the things people tell you about AI and its capabilities are just plain BS. And yet, I use AI every day – and not just for small, fun stuff or coding (where we know that it’s pretty good already), but as an assistant who never tires, who never gets annoyed by me asking it to do the same thing over and over again, and who, more often than not, delivers a unique approach to a problem. But I also use it as a unit of one – I inherently don’t trust its output, double-check and verify things it tells me, and use its output merely as an input into my own work and thinking rather than as an end product. This is a pattern that doesn’t scale – I catch AI making mistakes so often that I would (at least for the time being) never let it do things autonomously and unchecked; hence, I wouldn’t scale it beyond my own use. Which brings me back to Fitzgerald – the important bit in his quote is not just the ability to hold two opposed ideas in your head at the same time – but to keep functioning (well)…
And now, this…
Headlines from the Future
Dog Eats Dog. The background is a little nerdy, so bear with me. Tailwind CSS is a widely popular framework to design web pages – and a darling of AI code generators (there are specific reasons for that, outside of sheer popularity, but that doesn’t matter here). Chances are, if you ask ChatGPT, Claude, Gemini, or any other AI to create a website for you, it will use Tailwind CSS to style the page. A few days ago, the founder of Tailwind posted that his company had to fire 75% of its staff due to an 80% drop in revenue – caused by AI.
The company behind Tailwind makes money when people using their framework come to their website for help and documentation and then subscribe to their paid plans and services. Only, if you ask AI to build your website, you never go to Tailwind’s website…
AI will scrape your project site, users will never visit it for documentation and will never know about your commercial product.
Maybe one of the most direct idiosyncrasies of our AI-driven glorious new world. Dog eats dog.
━━━━━
Super Practical Advice on How to Implement AI. Most of the stuff you read about AI and how to adopt it in your organization is either so high-level that it’s useless, so specific and singular that it’s equally as useless, or simply AI-hype slop. Will Larson, CTO at Imprint (a FinTech company), has put together a blog post which is actually useful. It is highly recommended reading for anyone trying to figure out how to – actually – implement AI in their organization.
Given the sheer number of folks working on this problem within their own company, I wanted to write up my “working notes” of what I’ve learned. This isn’t a recommendation about what you should do, merely a recap of how I’ve approached the problem thus far, and what I’ve learned through ongoing iteration. I hope the thinking here will be useful to you, or at least validates some of what you’re experiencing in your rollout.
━━━━━
Commerce Disrupted. Our friend Jason “Retailgeek” Goldberg delivered the closing keynote at the National Retail Federation’s big conference last week. And when Jason speaks, we listen. Lucky for us, my former boss and blogger extraordinaire, Scot Wingo, was in the audience and recorded Jason’s talk. Jason was also kind enough to share his slides. If you are in retail/ecommerce, you have to watch this – and subscribe to Scot’s newsletter “Retailgentic.”
What We Are Reading
Trump May Be the Beginning of the End for ‘Enshittification’ – This Is Our Chance to Make Tech Good Again There is a glimmer of hope that could make technology good and fair again – and surprisingly, we might have investors and national security hawks to thank. @Jane
Anthropic’s New Cowork Tool Offers Claude Code Without the Code The new tool aimed at non-technical users is built for non-coding tasks, but it comes with its warnings – it can take strings of actions without user input and edit or delete files. @Mafe
Meet the Tree That Shoots Its Seeds at 150 MPH The sandbox tree solves a core evolutionary challenge by converting built-up tension into ballistic movements – I’m picturing shotgun seed dispersal! It is impressive how nature uses mechanics to gain a competitive advantage in dense ecosystems. @Kacee
Over 100 Episodes of Classic Sesame Street Have Arrived on YouTube This is too good not to share. I grew up on Sesame Street (the classic ones) – every kid should grow up on Sesame Street – and if nothing else, your dog might enjoy it when you leave him alone at home. @Pascal
Down the Rabbit Hole
💻 Logic (and Moore’s Law) dictates that our computers ought to get cheaper every year (or, at least, cost the same but become more powerful). With AI datacenters gobbling up anything from powerful graphics cards to RAM, this trend has been reversed: CES 2026 proved the PC industry is hosed this year
🚰 Talk about which – if you thought that energy is the big problem with AI datacenters, you should add “water” to your list of concerns: America’s AI Boom Is Running Into An Unplanned Water Problem
🔆 While the US is banking on fossil fuels, China is marching (actually, sprinting) towards a green future. Here are photos capturing the breathtaking scale of China’s wind and solar buildout.
👨🏼💻 Addy Osmani, Director at Google Cloud AI, shared his LLM coding workflow going into 2026 – super helpful for anyone who’s doing any coding with AI.
🦁 Some good food for thought: Tom Renner argues in “LLMs are a 400-year-long confidence trick” that LLMs are designed to exploit our cognitive biases and pull a long-standing confidence trick on us.
🧑🏼🌾 More evidence that AI’s productivity gains are nowhere to be found. And, at the same time, jobs lost to AI might be lost for good. This time, it’s coming from Forrester’s principal analyst JP Gownder: AI may be everywhere, but it’s nowhere in recent productivity statistics.
🤦🏼 I’ll spare you the “told you so” trope, but I have to admit that it is pretty ironic to see the company which renamed itself to cement its central role in the creation of the metaverse – shutdown most of its VR business: Meta is closing down three VR studios as part of its metaverse cuts
📺 We follow Doug Shapiro for his insights on the future of media for a while now. Here is his current thinking nicely summarized: My Base Presentation Deck - January 2026
⌛ More of a public service announcement – but as we are still in the first few weeks of the New Year, maybe you want to make this one of our New Year’s resolutions: Start your meetings at 5 minutes past
Pascal is getting excited for his upcoming trip to the Southern most tip of Patagonia. A place where even the Internet doesn’t work…
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.
When it comes to AI, I've long been in on the F. Scott Fitzgerald quote. You cannot model paradox with binary compute... that that piece of chocolate cake is probably delicious but also probably not very good for you if you ate the whole cake. A most human of experiences.
Hence I've always had little faith that digital compute will get us to any "AGI". We humans are quantum, not binary. Perhaps my convictions will change as quantum computing emerges, but not until then.