ReplyGuys

This account is no longer in use. Follow @replyguysintel

@qweasdbtc We’re hoping for the account to be restored within a day or two — else we’ll simply start a new account for this venture and go through the org approval process on there instead.

@askreplyguys Any update?

The askreplyguys account was broken while waiting for organization approval (gold badge) and the name reverted to an old name. This is awful timing and we are currently awaiting appeal from Twitter. Thanks!

Fortunately we caught this in time before the @askreplyguys name was taken by someone else. Nothing has changed regarding the project, thanks for the support and we appreciate any help from X in resolving this as soon as possible, which we fully expect. Thanks again!

@alliekmiller I’ve been experimenting with something of a similar concept, although a little more convoluted – but this is a much more elegant and clever way. Thanks for sharing.

One of my favorite ChatGPT / Claude hacks: answer levels. Give a level 1, level 2, and level 3 answer. A level 1 answer is boring, blah, basic. A level 2 answer is more complete. A level 3 answer is more creative. Then I ask AI for a level 4 answer. Works extremely well. AI systems are incredible at pattern matching, so lean into it

@gregthomasbooks @HiFromMichaelV @tsarnick The davinci-002 completion model was developed as a drop in replacement. It isn’t the same as davinci but it’s the most accessible and closest thing to it.

@HiFromMichaelV @tsarnick I never used GPT-3 (Chat was 3.5). What was it like? I assume a lot more hallucinations than we get now.

Mira Murati: GPT-3 was toddler-level, GPT-4 was a smart high schooler and the next gen, to be released in a year and a half, will be PhD-level