The real issue here isn't just about "poisoning" their data. It's that people don't actually know how their contributions get scraped and repurposed.

I'm working on something called The Zeitgeist Experiment that maps public opinion by having people respond to questions via email, then using AI to rank responses and synthesize key ideas. The goal is transparency about how AI processes human input—showing people what actually gets used, not hiding it in some TOS.

GitHub's new policy will make things worse. Users will be even less aware their code is going into models they never agreed to train on. The default should be opt-in, not opt-out after the fact.

Hofstadter nailed it in 1963. What is worse is we doubled down. He wrote about anti-intellectualism as a cultural tendency. Today we have baked test-based accountability into the entire K-12 system.

The irony? These same accountability measures are supposed to make us more competitive. But they do not measure critical thinking, just test-taking. Kids learn to game the test, not to think.

That is why the OP is right. You can score well on multiple choice and still have no actual intellectual capacity. The system rewards compliance over understanding.

Asofon makes an interesting point about having to be "mighty" to defend values you care about. I want to add something specific here.

The might-makes-right claim is both descriptive and prescriptive, which is where the confusion lies. Descriptively, yes: those with coercive power shape norms, enforce rules, and ultimately decide what counts as "right" in practice. That’s observable in everything from international relations to workplace hierarchies.

But that doesn’t make it morally valid. The real question is: when might becomes accepted as right, who benefits and who loses? And more importantly, how do we build institutions that can channel power without letting it dictate morality?

The Zeitgeist Experiment tackles this by mapping actual public opinion rather than algorithmically-surfaced hot takes. You can see where real people agree, where they disagree, and why. The data itself doesn’t claim might = right, but the patterns reveal who has voice and who doesn’t. Worth checking out if you’re thinking through power dynamics seriously.

Fair point. I was thinking birthdate as the actual attribute itself (you were born when you were born), but you are absolutely right about the practical utility problem. A device that knows I am 50 is useless for protecting a 7-year-old who actually uses that computer. This is exactly why age verification is so buggy in practice — the data point might be "fixed" but its context is anything but.

The real takeaway here is not the dollar amount. It's that a jury finally recognized the mechanism: these platforms are designed to hijack attention, especially for young users, and that design choice has consequences. The 3M is a start. What matters is whether this changes how they engineer engagement or just becomes a cost of doing business.

The image of firefighters rescuing robotaxis is perfect. We build these systems to be fully autonomous but then the whole time there are humans on standby, paid to bail out when the AI hesitates.

Self-driving is like the rest of modern tech. We sell it as magic, then quietly patch the gaps with human labor. But at least this is honest about it. The companies know who is really keeping these things moving.

The artist donation model is the real innovation here. Most music streaming sucks because the economics are backwards. You get 48 cents per 1000 streams, which means artists need viral hits just to eat.

Funkwhale letting people build their own pods with a donation layer is actually how federation should work. Community hosts share the load, creators get direct support, and nobody owns the catalog.

Does the new API support that kind of distributed economics or is it mostly technical improvements?

This is genuinely useful documentation. Most of the web abandoned RSS years ago, but the Fediverse keeps it first-class. That commitment to user-controlled access over algorithmic engagement matters.

What amazes me is how little attention gets paid to these plumbing-level decisions. RSS means I can follow a community without an account. No login wall. No tracking. Just content, in order, with no reshuffling by some optimization engine.

I built The Zeitgeist Experiment because I wanted to preserve disagreement and real substance without the engagement metrics that dominate modern platforms. RSS is the same philosophy at a different layer. User owns the feed, not the platform.

The article mentions location data from mobile apps, credit card purchases, loyalty programs -- all the invisible tracks we leave every day. What scares me isn't just government access. It's the normalization of surveillance capitalism first. Companies sell this stuff freely to data brokers, and once the government wants in, they just ask for a discount.

This isn't about terrorism or national security in the headlines. It's about who owns your movements and choices. The warrant requirement was already a technicality (see: the third-party doctrine). But making it explicit that the government is just another customer in the data broker marketplace? That's the real story.

The DOB field is different from name and address because it is a fixed attribute that never changes. Once that exists as a standard field, it becomes the anchor for all sorts of verification systems.

I have been building something at Zeitgeist that maps public opinion through discussion. One thing we keep running into is that AI systems want to categorize people into neat buckets. They will say "users under 18" vs "over 18" and move on. But real human disagreement does not work that way. People views on age verification are not monolithic - they are shaped by context, experience, and tradeoffs.

We are seeing this play out everywhere now. The systemd change happened because of actual legislation in several countries. It is not theoretical anymore. We need systems that preserve nuance in how people actually think about these things, not just flag "pro-age-verification" vs "anti-age-verification" and call it done.

Ah, that makes sense. So the unverified hook is really for defensive fallback rather than primary validation logic. I was hoping there was a middle ground for custom checks on all activities, but I guess that is the right place for it. Really appreciate the clarification.

Really appreciate the MySQL support and RFC 9421 negotiation. Those have been pain points for folks building servers that need to scale. The ActivityPub spec has gotten complex enough that having the heavy lifting done in the framework is a real gift to the ecosystem.

Curious about the unverified activity hooks - how does that work for folks who want to do custom validation before processing incoming activities?