Yeah, that was an oddly useful little feedback loop. You'd see some bizarrely specific query bringing people in and realize there was an entire question people were asking that you'd never thought to write about.
Graham says the constraint is that becoming more powerful has to make things better for the customer, but several of the strategies he recommends are exactly how companies eventually acquire the ability to stop caring quite so much about customers
Yeah, the bird use case is charming, but it's also a pretty good demonstration of how much information is sitting there waiting to be extracted from sensors that were installed for some completely different purpose.
Security cameras becoming a 24/7 backyard biodiversity monitor is already great. The fact that the same system eventually detected a neighbor farting in the driveway somehow makes it perfect.
A career probably shouldn't choose the destination for you. The danger is that it gives such clear feedback about progress that it's easy to let it do exactly that by default
A lot of careers have very effective feedback loops for telling you that you're progressing, while being almost completely silent on whether the destination is one you actually want
A lot of the tension around generative AI comes from the fact that the abstraction layer is often presented as interchangeable with the underlying craft
There may be real misconduct here, but "some employees mishandled confidential information" and "OpenAI must stop developing AI hardware" are very different claims
The scary part is not "age checks today, secret police tomorrow". It is that once there is a convenient identity-verification layer for normal internet use, it will be very hard to keep it limited to the original purpose... Every future regulator, prosecutor, platform and "think of the children" campaign will have an incentive to expand it