In this issue: Using chatbots as personal coaches and expert advisors. AI workflow hacks for manuscript editing. Google’s new video generator Veo3 and the ease of creating misinformation. And why publishers need to rethink content for Gen Z’s AI-first search habits with Rheinische Post’s Margret Seeger.

What we’re talking about: Studies of varying quality show that more and more people are using chatbots as coaches. There are tons of examples of prompts where the chatbot is supposed to pretend it’s not just a professional in something, but a famous person, and this personality is then supposed to help with life decisions, or evaluate an investment portfolio, business idea, or whatever.

Apparently we’re craving authorities—there are made-up interviews with Albert Einstein and other gimmicks. The more famous the person, the more data in the training sets, the better it works. There’s already the idea of using a person’s knowledge as the foundation for a chatbot. All of Andrew Huberman‘s collected podcasts (I know, problematic) are accessible through this chat and discovery interface by company Dexa.

I’m just waiting for more podcasters or even journalists to adopt this model. It’s only a matter of time. Google is already showing an example and has created a digital copy of author Kim Scott—basically a chatbot that draws on her knowledge and her bestseller “Radical Candor: How to Get What You Want by Saying What You Mean.” Because nothing says “personal growth” like a bot telling you to be more honest with your coworkers.

Knowledge that hasn’t yet flowed into the models’ training, or that the models wouldn’t spit out as quotes because they don’t want to break copyright that blatantly. The idea of expert chatbots is so obvious. The risks of the chatbot spouting nonsense in someone’s name are obvious too. You can minimize this risk and do nothing, or just go for it. Kim Scott and Google are going for it. What’s the worst that could happen? (Famous last words, probably.)

What else I’ve been reading:

Hands on: You want to see what a bot changed in your manuscript, but you don’t want to go full Microsoft Word “track changes” nightmare? A little workflow hack for manuscript editing: When you paste text into a chat window and get an LLM to edit it according to specific rules, like killing filler words or axing clichés, how do you quickly see what actually changed?

The output is just text again. But ChatGPT, Gemini, and the usual language models can handle basic text formatting like bold using Markdown. Here’s an example:

Read the following text, mark all adjectives in bold, like **so**, and return the text, no explanations.

Easy way to track edits without diving into track changes hell or doing tedious side-by-side comparisons.

One more thing: Google’s video generator Veo3 is out, and I’m a bit jealous because Henk van Ess had a brilliant idea. He spent about 30 minutes creating a fake news video. The concept: “How easy is it to create convincing political misinformation using AI before lunch is over?”

Veo3 clips are currently seven seconds long, some subtitles are added, and to keep the deception from being immediately obvious, Henk had to crop some of the videos. The result is one minute of chaos in Springfield. This could terrify anyone working in fact-checking, if they’re not already, you know, terrified.

I was traveling over the weekend and didn’t really have time to mess around with Veo3. But right before that, I stumbled across a funny post on Threads where pipers.jpeg asked Google what do to as a pigeon, and Google actually gave tips for pigeons—like staying close to the flock, avoiding dangers, general pigeon best practices.

This gave me the idea to generate the Daniel Craig meme where he says “Ladies and Gentlemen, it’s the Weekend” but with a pigeon. So I uploaded a screenshot of the meme to Gemini, had it describe the image, swapped Craig for a pigeon, and gave this prompt to Veo3:

This is a medium shot of a pigeon, who appears to be the actor Daniel Craig, standing on a stage that is characteristic of the set for the television show Saturday Night Live.

The Subject: The pigeon is the central focus of the image. The pigeon is standing upright, facing the camera directly. The pigeon is wearing a black short-sleeved polo shirt that fits snugly, accentuating his muscular physique, particularly in his shoulders and wings.

The Setting: The pigeon is in a television studio. The lighting is theatrical, with a spotlight illuminating him against a more dimly lit background.

To the viewer's left: There is a bouquet of flowers with red and pink blossoms on a stand. Behind the flowers, a portion of a set piece is visible, possibly resembling a window or doorway with some neon lighting.

Behind the subject: The background wall has a distinctive design featuring a series of large, circular openings with glowing blue lights inside.

To the viewer's right: The area is set up for a musical performance. Keyboards, stands, and other musical equipment are visible, which is a typical layout for the musical guest on Saturday Night Live.

The overall impression is of a host or guest delivering a monologue or introduction on a live television program.

The pigeon says: "ladies and gentlemen, it's the weekend"

This is the result.

It’s not quite the weekend yet, but I’ll see you after the next one.