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

When the company optimizes by algorithm, it makes everything legible to the algorithm — measurable, modelable. And human judgment, which is illegible to the algorithm, comes to be treated as noise to be eliminated. But it is precisely that judgment that captures what the metric does not see.

There is an essay that describes how the algorithm sees the world — how a platform optimized by an algorithm comes to see everything through the lens of what the algorithm measures. And it points to a broader transformation: what happens when the entire company becomes an algorithm. When the company optimizes by algorithm, it makes everything legible to the algorithm — measurable, modelable. And human judgment, which is illegible to the algorithm, comes to be treated as noise to be eliminated. But it is precisely that judgment that captures what the metric does not see.

Start with the idea of legibility. An algorithm can only optimize what it can read — what is measurable, modelable, reducible to a metric. To optimize the company, the algorithm needs to make it legible: translate everything into metrics it can read and optimize. That translation into legibility is powerful — it allows the optimization — but it has a cost: what is not legible to the algorithm, what does not fit into a metric, stays outside its vision. When the company becomes an algorithm, it makes everything legible to the algorithm, and what does not let itself be made legible — human judgment — stays outside what the algorithm sees.

Here is why human judgment is illegible to the algorithm. Human judgment — intuition, the perception of context, the reading of what is not in the metric — does not reduce to a measure. It captures what is hard to measure: the weak signal, the context, the exception, what the metric does not capture. By its nature, judgment is illegible to the algorithm — it does not fit into a metric the algorithm can read and optimize. And because it is illegible, the algorithm does not see it as value; it sees it as noise — a non-measurable interference in the optimization of the metrics it reads. Human judgment, illegible, is treated as noise precisely because it does not fit into the legibility the algorithm requires.

Here is the error of treating judgment as noise. Human judgment is not noise; it is what captures what the metric does not see — the weak signal, the context, the exception, what is outside what the algorithm reads. When the company, becoming an algorithm, treats judgment as noise to be eliminated, it removes precisely the capacity to capture what the algorithm cannot see. It becomes blind to what is outside its metrics — to what the illegible judgment would capture, but that the algorithmic optimization discards as noise. Eliminating judgment as noise does not clean the company of interference; it removes its defense against what the algorithm, by its limited legibility, does not see.

Notice the connection to the crisis outside the spreadsheet I pointed to. I pointed out that crises come from what is outside the spreadsheet — from the unmodeled, from what the metric does not capture. Human judgment is precisely the defense against that: the capacity to capture what is outside the spreadsheet, outside the metric, outside what the algorithm reads. When the company becomes an algorithm and eliminates judgment as noise, it removes that defense — it becomes exposed to what is outside the spreadsheet, because it eliminated the human faculty that would capture it. The company-algorithm is optimized for what is legible and blind to what is illegible — fragile precisely to what comes from outside the metric, which the discarded judgment would have captured.

See the paradox of algorithmic optimization. The more the company optimizes by algorithm, the better it gets at what is legible — at the metrics the algorithm reads and optimizes — and the more blind to what is illegible — to what judgment would capture. Algorithmic optimization improves the measurable and eliminates the non-measurable, making the company excellent at the legible and fragile to the illegible. The paradox is that the pursuit of optimizing everything, by making everything legible and eliminating illegible judgment as noise, creates a blindness to what is outside the metric — a fragility to the unmodeled that grows the more the company algorithmizes. The company-algorithm is more efficient and more blind at the same time.

It is fair, in balance, to recognize the power of the algorithm and that not all judgment is valuable. The algorithm is powerful — it optimizes the legible better than the human, and there is much value in that. And not all human judgment is signal; part is indeed noise — bias, error, inefficiency. The point is not to reject the algorithm or romanticize judgment, but to recognize that human judgment, illegible to the algorithm, captures what is outside the metric, and that eliminating it entirely as noise removes the defense against what the algorithm does not see. Maturity is combining — using the algorithm to optimize the legible and preserving human judgment to capture the illegible, instead of treating all judgment as noise to be eliminated in the name of optimization.

For the investor and the manager, this suggests distrusting the company that eliminated human judgment in the name of algorithmic optimization. The question about a company-algorithm is not only 'how well does it optimize its metrics?', but 'did it preserve the human judgment that captures what is outside the metrics, or did it eliminate it as noise?'. The companies that eliminated judgment are excellent at the legible and blind to the illegible — fragile to what comes from outside the metric; those that preserved it keep the defense against what the algorithm does not see. Whoever evaluates the company only by the optimization of the metrics misses the blindness the elimination of judgment creates; whoever recognizes the value of illegible judgment sees the fragility of the company that treated it as noise.

The rule of this moment: the algorithm optimizes what is legible and treats human judgment, illegible, as noise to be eliminated — but it is precisely that judgment that captures what is outside the metric, and eliminating it removes the defense against what the algorithm does not see. When the company becomes an algorithm, it gains in the legible and becomes blind to the illegible. Whoever treats judgment as noise optimizes the measurable and becomes fragile to the non-measurable; whoever preserves judgment keeps the faculty that captures what the metric does not see.

When the company becomes an algorithm, human judgment becomes noise. The company optimized by algorithm makes everything legible and treats judgment — illegible, not measurable — as noise to be eliminated, but it is precisely that judgment that captures what is outside the metric. Mark the transformation of the company into an algorithm not as pure optimization gain, but as a dangerous trade — the demonstration that making everything legible to the algorithm eliminates the illegible human judgment that captures what the metric does not see, that eliminating it as noise removes the defense against what is outside the spreadsheet, and that the company-algorithm, excellent at the legible and blind to the illegible, becomes fragile precisely to what comes from outside the metric — to the weak signal only judgment, discarded as noise, would have captured.

Leo Bentier