When cameras are installed, they issue many violations in the first month, followed by a steep decline in the second month. This suggests that the cameras effectively deter speeding by changing underlying driver behavior rather than relying on sustained revenue extraction.

NYC’s speed cameras are remarkably lenient: violations carry a flat $50 fine with no points (compared to hundreds of dollars and 3-11 points for police-issued tickets). They are also limited to school zones. Even so, we observe large safety benefits. 🧪 www.pnas.org/doi/abs/10.1...

Or, hear me out, we could mandate ignition interlock devices for 1/10,000th the cost and achieve similar effects on the whole.

This is the type of problem cost/benefit analysis was made for, yet everyone behaves like the only useful improvement is the one that costs a trillion dollars to execute.

Sort of, in the same way that people seem not to understand the things that their salary depends on them not understanding. If people accepted that LLMs are more mirrors than windows, that would obligate them to change their behaviors in ways they just don't want to.

"Dark Patterns" is the name of the UI work itself. Good resources here: www.deceptive.design

There's a version of Gell-Mann amnesia propelling the "what jobs will AI replace" discourse. Predicting whether a tool you (maybe) understand will replace a job you don't understand at all is obviously a ridiculous and unserious exercise. And yet...

The internal term for this is Juicing the Metrics. Nudge taught a whole generation of PMs the power of hijacking a default behavior.

I've always loved the phrasing, "why should I be bothered to read what you couldn't be bothered to write?"

But why deny yourself?

Whole development cycles have been lost to the AI fever dreams of investors, with nary a true customer request driving the roadmap.

As a CEO, I push my team *away* from AI at every opportunity. Some of that is because as a customer, I feel abused by AI platforms. I've spent four years building on Retool, for instance, and I'm ready to leave because a year ago they traded bug squashing for agent building and it's all gone to shit

I don't know how to convince well-meaning parents that elementary school education is not destiny (even college isn't!), but I do know that catering to their insecurities is a bad way to go about improving their kids' education.

I'm a product of "gifted" education and I sat down with my (now 2nd grade) "gifted" kids' teachers in Kindergarten and explicitly asked them to focus on social and emotional learning, because curiosity would take care of the rest. Kids (especially gifted kids) need to learn to be humans first.