What managing a team of Interns is actually like New

I’ve been working with Claude in my research platform on a 20+ page paper on the last ten years of AI, with Bloomberg as the backdrop to anchor the stories. We are using the work of four prominent AI figures to guide the insights. There are about a dozen diagrams, and almost as many sections. It covers AI terminology and business jargon. It’s complex writing.

And I’m glad to have my team of Interns, with Claude/Opus powering the bulk of the work — finding sources, distilling out information, producing charts and diagrams, and keeping it all in sync. Verifying that the links all work and the quotes are all accurate.

But like most teams of interns, it is a wrestling match with the actual writing quality. One problem I kept running into is that The Interns (Claude) was stuck on hedging many, many things about what we saw.

The hedging made “sense,” but it wasn’t always helpful. For example, the papers did NOT say anything about how they measured accuracy. None of them did. But this was like a drum beat in the writing. I honestly don’t find it surprising that they don’t, because I’m a human and have had to write public documents. If you hold up a measure in public then you can get held to it — sometimes legally.

In this paper it was coming off as an AHA! sort of conspiracy, which just isn’t the case. And it kept putting them back in. These callouts made the paper impossible to take seriously, because it was constantly redirecting the ideas toward what the Interns could NOT see or could NOT find.

That is what managing a team of Interns is like.

There were multiple tricks I found that helped, and we’re carrying that into skills work to get better in the future.

At the end of the thread I asked it to reflect on what it learned and write it in its own voice — a message from the AI writers, from the Interns. In their own words, unedited and only lightly coached.

Enjoy The Interns/Claude/Opus reflecting on how it could write better.