Personality is not a system prompt.
We’re on a mission to eradicate AI slop.
. . / \ __..-' '-..__ _.-' .-' '-. '-._ / .--' . . '--. \ ( ( .-' '-. ) ) \ '._ _.-' '-._ _.' / '._ '--..__..-' .------------. '-.__..--' _.' '--..___.' / \ '.__..--' .-' / o o \ '-. .-' .: | / | :. '-. .' .::: | \ / / | :::. '. / .::(@)::. \ '------' / .::(@)::. \ ; .::::::::::.. '.__ __.' ..::::::::::. ; _..-| ::: .--. ::::::. | | / .:::::: .--. ::: |-.._ .' .--| :: / () \ ::: .::. | |/ .::. ::: / () \ :: |--. '. / .' ; :: '----' :: /@@\::. | / .::/@@\ :: '----' :: ; '. \ | | / .::::::::::. \__/:::: |/ ::::\__/ .::::::::::. \ | | \ '._/ .::: .---. ::::::::::: ::::::::::: .---. :::. \_.' / '.__ .::: / .-. \ :: .-::::::::::::::-. :: / .-. \ :::. __.' / ::: ( (@) ) ::/ .--.::::::.--. \:: ( (@) ) ::: \ / .:::: \ '-' / :| (OO) |::::| (OO) |: \ '-' / ::::. \ ; .:::::: '---' :::\ '--' /::::\ '--' /::: '---' ::::::. ; | :::: .-::::::::::'----' .::. '----'::::::::::-. :::: | ; ::: / .--. ::: .--. :: /oo\ :: .--. ::: .--. \ ::: ; \ :: | ( \ :: / \:: \__/ ::/ \ :: / ) | :: / .---\:: \ '--' .' '::::::' '. '--' / ::/---. / .--. '::. '---' .--. '::' .--. '---' .::' .--. \ | | \ '::::.___.-' / \ :: / \ '-.___.::::' / | | \ '._ '._::::::_.-' | :: | '-._::::::_.' _.' / '._ '--. '----' _ / .::. \ _ '----' .--' _.' '--. \ .-' ) / / :: \ \ ( '-. / .--' \ \ / .-' / .' :: '. \ '-. \ / / | | ( ( .' .' :: '. '. ) ) | | __.' / \ '---' / :: \ '---' / \ '.__ '.__.' '.__.' / \ '.__.' '.__.' ( /\ ) \\// \/
The Problem
Mission: Prevent 1984.
Our consumption of AI is shifting our collective language towards AI-isms, whether we like it or not, and it's up to us to determine whether those AI-isms are hidden gems of the English language or lowest common denominator tropes that pull us down as a species.
The SFT objective — the loss function behind the problem:
LSFT = −∑t log pθ(yt | x, y<t)
SFT installs the voice, RLHF amplifies it.
Research
What is slop? To find out, we catalogued 'AI-isms' across the levels of linguistic taxonomy.
"This isn't just a productivity hack. It's a philosophy." negative parallelism
"the model — trained on public data — held up." parenthetical dash
"Choosing between Postgres and MySQL is an important decision for many engineering teams." CARS move 1 (Swales)
"Great question! This really gets at something fundamental." face-flattering act
Vision
-
Token-accurate personas and digital twins
Every AI persona you’ve ever interacted with has been an Assistant. We’re creating a world with believable personas and token-accurate digital twins.
-
Restore the loss of linguistic diversity for post-trained model outputs.
imagining a world where AI augments our use of the English language to communicate information, rather than constraining it.
Team
Built by Cody Hergenroeder and William Wang.
Contact: hello@sloplabs.tech