←News detail

Dex HorthyUpdated Oct 30 14:25

Dex Horthy reflects on his 2024 judgment of model capabilities

AI summary

Dex Horthy looks back at his 2024 claim that models are better at collecting, formatting, summarizing, and passing along context than at writing code and doing math, and now finds that statement a bit foolish. He points out that models are already solving Millennium Prize problems and cloning Redis, while for most people the worst part of the day has become reading unreviewed, lengthy AI-generated Slack messages and documents.

Body

In 2024 I said something which in retrospect, sounds kinda dumb

> we keep trying to get models to do this deeply technical stuff like write code and do math, but what they’re really good at is collecting/formatting/summarizing and relaying CONTEXT - we should be figuring out how to reduce the overhead of recording / relaying everything happening across a business

And now

1) models are solving millennium problems and cloning redis

2) the worst parts of most people’s days are reading verbose, AI-written slack messages and docs that the sender didn’t bother to review

Anyways keep learning yall

Source

Dex Horthy