"A Billion Nazis at the Table" - The Fediverse model proves contextual moderation by real humans is both easy and affordable. The presence of Nazis on corporate social media implies at least a tacit a
"If you’ve ever hosted a potluck and none of the guests were spouting antisemitic and/or authoritarian talking points, congratulations! You’ve achieved what some of the most valuable companies in the world claim is impossible."
Corporations don’t aggressively moderate and ban Nazis on their platforms because it would measurably negative affect their MAU stats, which is one of the primary metrics social media corps report on how “good” (read: profitable) their social network platform is.
Meta et al. will NEVER intentionally remove users that push engagement numbers up (regardless of how or what topics are being engaged) unless:
they determine it’s more profitable/less harmful to their business to do so
Which is another way the fediverse is better: The success metric is a vibrant, happy community, not MAUs or engagement numbers, so they make decisions accordingly.
Not to mention that because the fediverse doesn't require the collection of analytics it is less expensive to run. Most of the servers at Facebook are used to gather, sift, and deliver usage metrics. Actually serving content is a cheap and largely solved problem.
Twitter has always encouraged gawking at horrible behavior, and its culture has norms like "ratio" which promote "bad examples" so that they can be publicly shamed.