I think the real question is trust. Once a company starts working with the Pentagon, people will naturally assume the capabilities will expand over time. Even if today’s contract has limits, those limits can always change later.
What makes the Gervais Principle interesting is that it frames organizations as systems that are supposed to become pathological over time. Instead of reforming them the economy just replaces them through mergers, layoffs, and new startups.
In that sense, the real “governance mechanism” isn’t internal management at all but the external Darwinian churn of the market.
It survives because it already solves the basic problem well enough. You can reach almost anyone and communicate without needing to join a platform or build a network first.
It’s simple, open, and relatively low-pressure compared to most modern communication tools. A lot of newer products try to improve it, but often they’re mostly adding layers rather than solving a fundamentally missing capability.
What I probably need most right now is honest feedback. I’ve been building something and I’m trying to understand if the idea actually makes sense outside my own head.
The Belmont analogy is great, but the deeper point is even scarier: most of the industry is giving non-deterministic systems direct access to deterministic infrastructure (databases, shells, email, etc).
Historically we spent decades reducing automation privileges and adding layers of verification. Agents seem to be reversing that trend almost overnight.
I'm building a human-curated map that organizes people and sources by topics (health, skills, business, mindset, etc.). Everything is categorized manually, no AI classification. Started as spreadsheets and now turning it into a prototype.
The uncomfortable part is that they try to solve a real problem (protecting minors) by requiring universal identification. In practice this means every adult has to prove who they are just to access any part of the internet. Once that infrastructure exists, it’s hard to imagine it not expanding beyond its original purpose.
This is fascinating. One question I always have with large personal datasets like this: at what point did you start getting genuinely surprising insights versus confirming things you already suspected?
380k datapoints sounds incredible but I imagine the real challenge is turning that into decisions that actually change behavior