For clarity, the whole point of this launch is to enable people to easily customize their own policies so that we can support a plurality of content classification perspectives online! It is actually a solution to the problem Evelyn highlights in that piece.

#MakeMasnickWrongAgain

Social media algorithms push people who are near the extremes to further extremes. But it doesn't have to be this way.

The standard damage control post from tech companies in these situations is to implicitly put the responsibility/blame on users. We can (and should) debate whether OpenAI's promised mitigations are sufficient, but the fact that they are tackling the *product problem* head-on is a vital first step.

Yes, would definitely love to directly integrate with roost/coop. Let's chat!

Would love for you to try plugging in zentropi next time! Can just point Claude at our documentation at zentropi.ai/api and it should just work after you generate an API key. I've actually done this myself for a Bluesky feed prototype.

Reasonably sized PBCs FTW.

We've done some benchmarking on how well gpt-oss-20b does on classification tasks. Significant step up from gpt-4o, especially for normie policies, but still stubbornly inflexible for custom policies.

When I access that @nytimes.com article, it doesn't say "for Facebook's culture". Instead it says "for an inclusivity initiative at Facebook that encouraged employees’ self-expression in the workplace". Was it edited or updated?

@caseynewton.bsky.social In terms of other news feed demotions that could be rolled back (or maybe already have been), take a look at this blog post. Would be smart to track it as it is frequently updated: transparency.meta.com/features/app...

A huge chunk of demoted misinfo on FB are not explicitly about politics at all. Now these hoaxes will get massive amplification. One recent example from Snopes (which was an important fact checker!): www.snopes.com/news/2025/01...