DooDooLamb News
Nobody Knows Anything - Derek Thompson
Brief published March 1, 2026 · Original source published February 28, 2026
Original reporting by maxbarners at derekthompson.org.
Automated brief. Verify important details at the original source.
Silicon Valley's Newest Hobby: Cosplaying Economic Prophets While Building ChatGPT Wrappers
The tech industry has officially entered its "guy at the bar explaining cryptocurrency" phase, except now everyone's an AI economist and the bar is Twitter. This week brought fresh evidence that the entire conversation around artificial intelligence's economic impact has devolved into a marketplace where venture capitalists sell competing science fiction novels, and somehow we're all supposed to pretend these are quarterly earnings reports.
Derek Thompson, writing for his newsletter with the kind of clear-eyed analysis that makes you remember why actual journalism exists, dropped a reality check so hard it probably caused a few AI startups to pivot to "stealth mode." His central thesis cuts through the noise like a debugger through spaghetti code: nobody actually knows what AI is going to do to the economy, but everyone's acting like they've got the next five years mapped out in a spreadsheet.
The comparison Thompson draws to Orson Welles' 1938 "War of the Worlds" broadcast hits different when you realize we're living through our own version. Except instead of Martians landing in New Jersey, it's ChatGPT landing in every corporate PowerPoint, and instead of people fleeing their homes in panic, they're fleeing to Medium to publish their hot takes about the coming jobpocalypse. The radio audience in 1938 at least had the excuse of tuning in mid-broadcast. Today's AI discourse participants are choosing to walk into the theater, sit down, and then act shocked that the aliens look suspiciously like actors in costumes.
What makes this particularly delicious is how the uncertainty plays out in practice. On one side, you have the "AI is going to automate everything and we'll all live in a post-scarcity utopia" crowd, typically populated by people who've never had to explain to their parents how to unmute themselves on Zoom. On the other, you get the "AI will steal every job and leave humanity obsolete" faction, usually led by the same folks who thought the internet was just a fad for academics. Both camps share one crucial trait: they speak with the confidence of someone who just discovered they can ask ChatGPT to write their quarterly planning documents.
The economic data Thompson highlights reads like a greatest hits album of "things that make economists develop drinking problems." Productivity measurements are all over the place, companies are reporting cost savings that somehow don't show up in their actual spending, and everyone's favorite metric (whatever makes their PowerPoint look better) changes depending on who's presenting. It's like trying to measure the impact of smartphones in 2008 by counting how many people are using them to make phone calls. The real revolution happens in the spaces between the metrics, in the workflows nobody thought to track, in the problems that get solved before anyone realizes they were problems.
Meanwhile, the actual builders – the engineers shipping features, the designers crafting interfaces, the product managers trying to figure out what customers actually want – are caught in the crossfire between breathless prophecy and grinding reality. They're building RAG systems (Retrieval Augmented Generation, which is basically teaching AI to cheat by looking up answers in real time, like having Wikipedia open during an exam) and fine-tuning models (the AI equivalent of teaching your autocorrect that you really do mean to type "ducking" sometimes), while the discourse around them oscillates between "this will replace all knowledge work" and "this is just autocomplete with delusions of grandeur."
The venture capital ecosystem has adapted to this uncertainty by doing what it does best: throwing money at the problem and hoping someone figures out the business model later. Every pitch deck now includes a slide about "leveraging AI to disrupt legacy workflows," which is startup speak for "we added a chatbot and hope that's worth $50 million in Series A funding." The result is a Cambrian explosion of companies building essentially the same thing – wrapper APIs around OpenAI's models – while claiming they've invented the future of work, customer service, content creation, and possibly human consciousness itself.
Thompson's piece lands hardest when he points out that this isn't just harmless speculation. Real decisions get made based on these competing science fiction narratives. Companies restructure around AI capabilities that exist more in press releases than production systems. Workers spiral into existential dread about jobs that might get automated by technology that might not actually be able to do their jobs. Investors pour billions into solutions for problems that might not exist while ignoring actual problems that definitely do.
The reality check here isn't that AI doesn't matter – it clearly does, in ways we're still discovering. The reality check is that admitting uncertainty doesn't make you less smart; it makes you more honest. The companies building genuinely useful AI tools tend to be the ones talking least about revolutionizing everything and most about solving specific problems. They're improving code completion, making customer service chatbots that don't make you want to throw your phone, and building search interfaces that actually understand what you're looking for. Revolutionary? Maybe. Worth writing dystopian fan fiction about? Probably not.
The funniest part of Thompson's analysis isn't the critique of the hype cycle itself – anyone who lived through web3, mobile-first, cloud-native, or any other technological buzzword casino knows this song by heart. The funny part is how quickly "nobody knows anything" became the most radical position you could take in Silicon Valley, where admitting ignorance is apparently more controversial than claiming you've solved artificial general intelligence.