After staying off Twitter reasonably successfully for years, I cross-posted my AI Tweets there a few weeks ago. I had something very Twitter-shaped to say, and I thought it was important to get out, so I do think this was worth it. And it all went well: none of this is complaining about the comments I got there.

Coming back a few times to check notifications, however, it's been very good at baiting me: Tweets that are confidently wrong in cases where I have relevant and uncommon knowledge. The pull to dive in and share what I know is very strong! Then this bleeds over to the far broader case where people are wrong, and you have a large potential time sink.

If it were just the time sink, I'd stop resisting. I spend some time on HN and Reddit, and to the extent that Twitter could substitute for that by showing me things I was more interested in, that wouldn't be an issue. The real problem is the culture.

Twitter has developed a culture of raising the temperature, and generally being mean. People dunk on each other, and are aggressively uncharitable. I see how the incentives push this way: fights and taking sides draw people in, and an engagement-based algorithm [1] will move the crowd in this direction. You internalize the drive to make the internet numbers representing your influence go up, and I know I'm very susceptible to this. But the practices this develops are bad for discourse, and becoming "good at Twitter" would strengthen harmful aspects of my personality and push me towards being a worse person. [2]

A lot of people swear by heavy use of blocking, lists, and the "Following" feed. I expect this to help, but not enough: the bait is in the replies and quote-tweets, and from the outside it looks like the culture seeps everywhere.

This makes me a big fan of Zvi's roundups (example), where he quotes relevant Tweets and gives context. Just as academic AI publications have moved from journals to preprints, a lot of the news has moved to Tweets. And AI news is both really important and poorly covered by traditional media: if Zvi didn't put out these roundups I might have to break down and read Twitter. I'm super thankful that this means I can stay informed without opening myself up too much to Twitter culture.

So I'll continue to accept that I'm missing out on a place where people have interesting public discussions on topics I care a lot about: engaging would expose me to an incentive gradient I don't think is worth the risk.

[1] Now that we have better algorithms, in the form of LLMs, we could have much better ranking signals. For example, you could write an equivalent of the HN guidelines or Duncan's Basics of Rationalist Discourse and then give more visibility to posts that did a good job of exhibiting the kind of discourse you'd like to see more of. Of course this requires the platform to want this: it likely lowers engagement, and hence revenue. But maybe there are paths that run via taking a long view on building a site that's more influential or where people feel better about the time they spend there.

[2] There are of course many people who remain wonderful people despite being popular on Twitter, through natural immunity or impressive resolve.