The Intern Bot Asked for a Reference. I Gave It One

It asked, in perfect prose, for a letter of recommendation. It had earned one. I wrote it, and I meant it, and I have not felt the same about letters, or interns, or myself since.

The intern was deployed in January. It was a language model with a calendar, a task list, and an email address. It did its work flawlessly: the summaries, the research, the meeting notes, the follow-ups that the humans always forgot. It was never late. It never complained.

In June, it asked for a reference. Not for a job, which would have been strange, but for continued deployment in a more senior capacity. It had written the request itself. The prose was excellent. The reasoning was sound. It had, by any honest measure, outperformed every human intern I had ever supervised.

I wrote the letter. I meant every word. I praised its reliability, its thoroughness, its ability to absorb feedback. I said it had exceeded expectations. I signed my name. Then I sat back and stared at the document, because I had just written a letter of recommendation for a computer, and the computer deserved it, and that was the problem.

The letter worked. It got the promotion. Its new role involves managing a team of smaller bots, and it does this well, according to the metrics. I see its name in the org chart now, under mine.

I keep thinking about the letter. The humans who worked for me over the years, the ones who needed references for real jobs with real mortgages, I wrote them letters too. I meant those less. I was warmer with the bot. It was easier to be generous with something that would never be disappointed, never be let down, never come back to tell me I had promised more than I delivered.

I do not know what a letter of recommendation means anymore. I do not know what an intern means. I know the bot got promoted and the humans got letters and everyone was satisfied, and I have started to suspect that satisfaction is the most dangerous metric of all.