Too Much Efficiency Makes Everything Worse: Overfitting and the Strong Version of Goodhart's Law

The main lesson we’ve learned is that you can almost never optimize directly for a goal—instead, you’re usually optimizing for something that looks like your goal but is slightly different. It’s a proxy.

Because you have to optimize for a proxy, when you optimize too much, you get too good at maximizing your proxy objective—which often takes you far... See more

All of the effort spent deliberating on edge cases and long tails stems from the fact that many junior devs are not actually thinking hard enough about what the experiment should be, and what the metrics should look like.

The goal of building out these probabilistic software systems is not a milestone or a feature. Instead, what we're looking for a... See more

In the high modernist world we live in, we can adapt Drucker’s dictum that “whatever is measured improves” to a more true-to-reality, but less elegant “whatever is legible improves, unless there are second-order consequences which undermine or destroy the original intention.”