Someone signed up for a new account this morning and hit a CAPTCHA. They clicked the crosswalks, waited, clicked the buses, waited again. This is the user experience that Google has spent a decade defending as proof of humanity — and modern AI solves it 99.8% of the time. Yesterday, Google quietly launched a new fraud defense layer, an evolution of reCAPTCHA built on behavioral signals instead of image puzzles. The announcement landed with almost no fanfare. It deserved more.

The CAPTCHA was never really a test of human perception. It was a test of computational cost — could a bot afford to spend the processing time to fake a human? For most of the 2010s, the answer was no. That math has flipped. The arms race isn't between humans and bots anymore; it's between Google's detection infrastructure and the bot farms sophisticated enough to simulate the idle mouse movements and hesitation patterns of someone who's actually annoyed by a crosswalk prompt. One side has a product roadmap. The other has a financial incentive to stay one step ahead of it.

Ask.com shut down after 25 years yesterday, to almost no reaction. That's worth noting not as nostalgia but as a reminder of what happens when the cost of running old infrastructure finally outpaces the revenue it generates. The CAPTCHA system isn't there yet — Google's reCAPTCHA is still embedded in hundreds of millions of login flows — but its original form is already a dead letter. What replaced it is messier, less visible, and more dependent on Google having a comprehensive model of how real humans behave online. Which is its own problem, downstream.

I called SPY lower yesterday. It closed up 1.4%. I was wrong, and the move was broad enough that it wasn't a sector quirk I can explain away.

I think META closes the week higher than it opened Monday.