When the company becomes an algorithm, human judgment becomes noise.

An operational reading of Eugene Wei's essay: organizations that force legibility destroy what they can't measure. And what they can't measure is usually what matters most.

Eugene Wei wrote about how TikTok functions as a learning machine — it observes user behavior, trains a model on what works, and delivers content calibrated to maximize engagement. Wei's central argument is that TikTok's algorithm is more honest than social graph-based networks because it measures what people actually do, not what they claim to prefer. It's an essay about technology. But the structure he describes is identical to organizations that force legibility — that convert everything diffuse, contextual, and judgment-dependent into metrics, processes, and rules because they can only manage what they can see.

The problem with companies that operate like algorithms isn't that they measure wrong. It's that they optimize the measurable at the expense of the immeasurable. You create customer satisfaction KPIs and start managing toward the number — not toward satisfaction. You measure productivity by hours or deliverables and start managing toward deliverables — not toward real impact. What falls outside the measurement system doesn't disappear. It keeps happening. But it stops being managed and starts being merely tolerated — until the accumulated cost of ignoring it becomes too large to hide.

Human judgment is what an organization has when it can't reduce a problem to a metric. It's what allows you to say this customer is worth more than the CLV indicates, or this hire is right even if the candidate doesn't score well. Companies that build systems to eliminate the need for judgment don't become more efficient. They become more rigid. And rigidity, in changing environments, is vulnerability. TikTok's algorithm works because the product is content and the feedback signal is immediate. In most companies, the product is complex and the feedback signal arrives too late for any algorithm to optimize before the damage is already done.

Leo Bentier