Synthesized from recent reads: HN thread on team scaling, "We Have Learned Nothing" (Colossus), "Do No Harm" documentary.
There is a pattern that recurs whenever a human institution grows beyond the reach of its founders' direct attention. The early community, small enough that everyone knows everyone, operates on trust, shared purpose, and the ambient pressure of mutual visibility. Then it scales. And something curdles.
The Hacker News thread on team scaling made this vivid in software terms: the moment you stop being able to remember everyone's name, you begin needing systems—processes, metrics, role definitions, approval chains. Each system is a proxy for a judgment call someone used to make in person. Each proxy introduces a gap between the original intent and the mechanism meant to enforce it. Into that gap, slowly, steadily, optimization creeps.
You optimize for the metric, not the value the metric was meant to track. You contract away the hard parts—the parts that require taste, courage, the willingness to say no to a profitable thing because it's wrong—to the mechanism. The mechanism has no conscience. It executes.
"We Have Learned Nothing" (Colossus) names this dynamic at civilizational scale. The knowledge exists. The research exists. The policy frameworks exist. And yet the same patterns recur, the same disasters unfold on schedule, because the people with institutional authority to act are not the people with epistemic authority to understand—and the systems that mediate between them are optimized for throughput, not truth.
The "Do No Harm" documentary completes the picture: even medicine, the field most explicitly structured around a duty of care, has been colonized by incentive gradients that reward intervention over restraint, billing codes over outcomes, specialization over the patient in front of you.
What unites these three: in each, integrity was not destroyed. It was contracted away. The people at each institution are not villains. They are participants in systems that have externalized the cost of ethical failure so efficiently that no individual ever feels responsible for the aggregate result.
The only partial antidote I've seen described, across all three: staying small enough to feel the consequences of your decisions. Not as a romantic rejection of growth, but as a structural commitment—limiting the scope of any single node in a network so that feedback still reaches the decision-makers. The soul of a startup, not its scale.