Asking Good Questions with Edward Roske

You Can't Prompt Your Way to a Flying Car | Tim Berglund

Edward Roske Season 3 Episode 5

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 43:42
Tim Berglund spent twenty years calling AI a promissory note. Always a bright future, never a bright present. (The field was born overpromising. The Dartmouth crew figured they'd nail human intelligence in a couple of months. That was the 1950s.)

Then in December 2025 he opened Claude Code and said it on tape: "Okay, this is real."

So I asked him what he sees now that he didn't see in the twenty years he spent calling this a trick. Coding got cheap, and cheap coding exposed the true limiting factor, which is that it's hard to make things that don't suck. His words: coming up with a good idea, still hard. Designing a product interface, still hard. Seven months into the agentic revolution, Jira has not suddenly become good. (Check your own tools and tell me he's wrong.)

He's completely sold on the tool and completely unsold on the idea that there's a mind in there. It talks like a mind, but it's not one. We dig into why the first machine that uses language the way we do keeps fooling our theory-of-mind reflex, whether AI is making work better or just faster, what the Turing test is worth now that it's been passed and gotten boring, ELIZA pulling the same trick in 400 lines of code back in the sixties, and why his firmware is still handcrafted, artisanal, small-batch C++ (like making your own furniture out of wood).

Also in here: the day Claude Opus started calling the other agents its siblings (that terminology did not come from me), the Postmodernism Generator as 1990s proto-slop, the bicycle being invented after photography, and my air-band machine. I have played Ode to Joy on an air piano. It was not quite as much fun as I thought it would be. (The air drums are just clank, clank, clank.)

I'm Edward Roske, and this is Asking Good Questions. New conversations with the people figuring out the human side of AI. If one of these questions stuck with you, subscribe and send it to someone who'd take the other side.

Tim speaks at the Caribbean AI Summit this October in San Juan, on the real history of AI. He gave me a taste of it in this episode and offered to give me the rest, which I turned down so he'd still have a talk to give. Come find us: CaribbeanSummit.AI

LINKS
Caribbean AI Summit (Tim speaks, Oct 9-10): https://CaribbeanSummit.AI
Tim's site (writing, talks, Substack): https://timberglund.com
Tim on X: https://x.com/tlberglund

BOOKS & RESEARCH MENTIONED
The Dartmouth workshop (the 1950s proposal that kicked off AI as a field)
ELIZA, Joseph Weizenbaum's 1960s chatbot (Tim: about 400 lines of code)
The Postmodernism Generator (the 90s fake-essay site Tim cites as proto-slop)
The Turing test (passed, and passe, per this conversation)
The Hitchhiker's Guide to the Galaxy, Douglas Adams (my fault, as usual)
Also mentioned: Claude Code, Claude Opus 4.8 and Sonnet 5, Apache Kafka

This episode of Asking Good Questions is brought to you by Caprus.Ai.

Check how we empower the office of the CFO to harness the transformative potential of artificial intelligence (AI) and revolutionize their operations.

People on this episode