Andor is more of a thriller. How does a rebellion against a state collapsing towards totalitarianism take shape, with Star Wars livery.

And Andor. Either before or after Rogue One.

I'm working up to it. A year ago I was comfortable letting AI autocomplete by debug prints. Today I'm comfortable letting AI do a change within a well-defined system, or helping me debug a difficult behavior. Maybe next year I'll be comfortable letting it build entire systems. But what will not change is that I always want to control the design, if not the implementation.

You always had to check it. The difference is before you explained the change you want to a junior dev and check it in 3 days. Now you give the same explanation to an AI and check it in 3 minutes.

I'm passing on secondhand explanations, I may have the details wrong.

Actually, in normal law, stall recovery is full power full nose up. Of course, they weren't in normal law...

According to an extensive discussion of this I found elsewhere, for much of the accident sequence the plane was SO nose high that the stall horn turned off because static and dynamic pressure were equal.

Whenever the nose went down a bit, the stall horn would come back on, producing exactly the wrong reaction----go back up.

Unfortunately, the situation was unrecoverable for several minutes before impact according to simulations.

There is a certain mindset that says, if the thing everyone is warning about hasn't happened yet, then it must not be a real problem. That's human nature; it's the same reason people ignore car alarms. But it's flawed.

What we should be doing is recognizing that if the predictions keep pointing the same direction, there's something they're all pointing at. Improving data may change how well we can tell how far away the thing is, and how large it is. But that there is a thing in that direction we should be concerned about is not changing.

I found a 2015 article linking Miami, sea level rise, and the year 2025. That one isn't saying that it will be underwater by 2025, only that the climate trajectory is such that its eventual fate will be sealed by 2025 if changes aren't made, because climate doesn't turn on a dime. I don't know if that's the one you're referring to, of course. But I'm sure whatever it is was making the best prediction it could given the data.

The difference between transit and taxis is the same as it always was. The difference between taxis and robotaxis is similar to the difference between taxis and Ubers: none, except technology.

That's why self-driving is not a fully solved problem, nor is it realistic to expect it to be solved all at once. But wow, it's much further along than it was a decade ago!

These solve a different problem than transit does. And while the loss of gig jobs is unfortunate, that argument has never stopped the march of technology. Personally, I want a self-driving RV. Go to sleep and wake up at the destination.

That AI enables a fundamentally different approach is not in question. That Google has positioned itself at the forefront of that is obvious. What remains to be seen is whether humans or TPUs have a higher price tag for the same quality.

Google is hoping for economies of scale to kick in. And they are, to an extent. The newest flash models are cheaper than the pro models a year ago.

The stock market doesn't think so. Goog is up 128% YoY.