Abstract: Online speech governance stands at an inflexion point. Platforms are emerging from the state of emergency invoked during the pandemic and lawmakers are poised to transform the regulatory landscape. The importance of what emerges from this moment can hardly be overstated: how platforms write and enforce the rules for what speech they allow on their services shapes the most important channels for communication in the modern era, and has profound consequences for individuals, societies, and democratic governance. Understanding how online speech governance arrived at this moment illuminates the tasks that the institutions created during this transformation must be designed to do. This history shows that where online speech governance was once dominated by the First Amendment tradition’s categorical and individualistic approach to adjudicating speech issues, that approach became strained and online speech governance now revolves around the principles of proportionality and probability. Proportionality requires governance to no longer focus on the speech interest in an individual post alone, but to also take into account other societal interests and place proportionate limitations on content where necessary. But the unfathomable scale of online speech governance makes the enforcement of rules only ever a matter of probability: content moderation will always involve error, and so the pertinent question is what error rates are reasonable and which kinds of errors should be preferred. Platforms’ actions during the pandemic have thrown into stark relief the centrality of these principles to online speech governance, but also how undertheorized they remain. This article reviews the nature and causes of this shift of online speech governance from a “posts-as-trumps” approach to one of systemic balancing, and what this new era of content moderation requires of platforms and their regulators.