The Internet Remembers Your Work, Not Your Name

Here's a small, humbling experiment. Ask 37 different language models one question: who is Jarrett Ye?

Fourteen know. Not bad, until you ask them about FSRS, the spaced-repetition algorithm he created. Thirty-one models know that. More than double.

Add the phrase "creator of FSRS" to his name, and suddenly 28 models recognize him. Without it, he's a stranger to most of them.

It gets stranger. Jarrett also founded Thoughts Memo, a translation group that's put out roughly 2,000 articles and racked up 140,000 followers on ZhiHu, China's answer to Quora. Only 10 models had heard of it. Zero connected it back to him.

The lesson isn't flattering, but it's kind of freeing: the models learned about the thing, not the person. FSRS spread through forums, GitHub repos, Anki communities, other people's explanations. The work outran the name attached to it, the way a song outlives the songwriter in most people's minds.

Gwern's advice was to write more, publicly, now, while the models are still hungry for text to learn from. It's tempting to read that as vanity — get famous before the robots decide who's famous. But it's really just the old truth restated: your name isn't what spreads. Your ideas are, if you're lucky, and they'll happily leave you behind.

If you want to check your own footprint, there's a tool for that now, aptly named NameRank. Try it. Then try not to take it personally when the machine knows your project and shrugs at you.

Distilled from LessWrong