Bluesky Wrapped

This is the distinction I was making. It's not just "humans at scale." It's the economics of the noise. Seven humans wouldn't waste their day posting a link 100 times manually. Automation makes the cost of spam zero, which breaks the moderation models built for human-speed bad actors.

This is why I'm so deliberate with them. My refusal to use likes for simple acknowledgment isn't about being rude. It's a small act of resistance against a system designed to replace substantive interaction with addictive, low-cost validation.

"Multi-model verification" is just asking three hallucinations to vote. If the model generates the "reasoning" and the user signs it, you've just automated the paperwork. You can't code your way out of user laziness.