In this issue: Defiant language models are reading our emails. The death of vibecoding. Books made out of thin tokens. And the radical change in audience expectations with The Telegraph’s Michelle Brister.
What we’re talking about: At least two newspapers in the US have printed a summer book list that was created, well, with the help of AI. And in a really annoying way: some of the books simply don’t exist. Made up.
There’s a lot in the story behind it, because the list wasn’t generated by some unknown person, but by someone who knows better. But was under time and cost pressure. And then didn’t pay attention. NPR’s Elizabeth Blair has the full story on what happened.
This Wednesday, Clare Spencer and I are talking about this – and other – AI fails in Berlin at re:publica. Clare has worked for the BBC, for a major British publisher, and now writes for Generative AI in the Newsroom. We have slightly different takes on this mess, so it should be interesting.
I’m not going to waste too many words here on Google’s I/O, the OpenAI billion-dollar deal with ex-Apple designer Jony Ive, or Codex, the autonomous coding agent. But I do have some links for you:
If only it were just a tool – but instead Claude 4 sometimes shows a stubborn mind of its own, complete with blackmail and snitching. Simon Willison read through the official documentation.
The Pulitzer Prize is once again highlighting investigations that use AI – but we’re talking “regular” AI here, not ChatGPT. NiemanLab has the details.
Okay, this is slick PR, but: Anthropic got Rick Rubin to do some visual code experiments based on Lao Tzu for the Claude release. Translation: with Claude access, you can vibe-code pretty animations.
ChatGPT stifles idea diversity in brainstorming, a recent study says. Heads up though: researchers were looking at ChatGPT 3.5, which dropped in November 2022 and has since been discontinued.
Google’s new AI chatbot mimics Perplexity and OpenAI’s ChatGPT, prompting questions about the future of information discovery. (Tripp Mickle, New York Times)
Dan Shipper of Every startup uses AI to augment, not replace, writers – a contrarian take amidst AI-fueled media disruption. (Benjamin Mullin, New York Times)
Google’s Veo 3 can spit out videos that are seconds long and look deceptively real like US television, at least at first glance. (Ethan Mollick, LinkedIn)
Calling it out: Apparently “vibecoding” is already burned territory in certain circles – too many LinkedIn thought leaders got their hands on it, and now it reeks of hustle culture BS. And anyway, where does it lead if we just do stuff with computers without deep understanding? Hippie dreams. It has to be hard, incomprehensible, and full of barriers to entry.
Of course not! Respected media hype researcher Johannes Klingebiel also finds writing local, problem-focused software with minimal prior experience useful. But since “vibe” now stands for ignorance and cluelessness, he’s looked for a replacement. He digs up a term from 2004: Situated Software. It comes from author and professor Clay Shirky, who uses it to mean small programs for a dozen people, for a specific problem. It’s not about scaling to millions – it’s about solving the thing right in front of you.
Work in progress: I’m still putting together a list of AI writing tools for a next issue. If you’ve got favorites that don’t make you feel like you’re contributing to the heat death of human creativity, send them my way. I’m thinking tools like LLM Peer Review or Lex.page.
One more thing: I recently broke my own “celebrate good things instead of punching down” rule and called out an AI tool that promised instant content creation based on trend signals. Called it a “dystopian hellscape,” because honestly, that’s what it felt like – another slop machine contributing to the internet’s ongoing content pollution crisis.
The Hacker News crowd was about as gentle as you’d expect (read: not at all). And then the founder reached out to me: “Our original ‘no research, no writing’ messaging was everything wrong with AI tools,” he said. “It promoted the exact lazy, soulless content creation that’s flooding the internet.”
Did THEFUTURE somehow make the world marginally better? The landing page does read differently now. But I can’t shake the feeling this is just another pivot in the endless startup hustle. The messaging changed, but did the product? The founder’s staging his lesson learned, but the hustle continues.
See you next time!
