There are other terminal image protocols too, iTerm and Sixel. Doesn't mean you have to use those terminal emulators, you can use one which supports one of those protocols.

This is something I really miss in std Rust. With all the zero allocation parsers out there, often I need to own both source and parsed data (pointing to the source). I've been using Ouroboros, Yoke sounded more scary. 😁

Thank you for sharing. It touches upon something I've been thinking a lot about: when it comes to teamwork, when it's better to ask a colleague for help, context, opinion, and work together, vs when I should use LLM and work on my own.

Why? Isn't it good nerdy stuff? Or do you mean that e.g. poisoning should belong to a speciality 3rd party crate and the stdlib should abstract away from this use case?

I like that. To take it further and to make it consistent with unwrap_or and friends, would eg unwrap_or_panic be more predictable than just or_panic? 🤔

I guess it's a new variant of the old joke. Rephrasing it (badly): Spotting that LLM hallucinates is the second worst thing. What's the worst thing? Not spotting it.

There are multiple contributors. Inefficient cars pollute, heavy cars damage roads, and large cars take up space. Heavy cars (including electric) have high microplastic emissions from tyres (plus subjectively, I find them very noisy.)

Yet another pointless invention. 😅 Why not look at other countries and adopt what has already proven to work?

I'm always surprised how much of good/reusable material gets thrown out by builders/contractors. Sure, it's cheaper/less work, but it still feels wrong. Personally, we tend to offer these on Freecycle (in the UK), it's amazing how much gets reused.

For me, purely from passenger's point of view, I prefer when lines have intersections. In principle, circle line adds these intersections so that I don't need to travel to hubs/centers,especially when the loop circles around busy centres. I'd rarely use the loop line itself to get to my destination.

I only have a very basic understanding how things work, but to me it feels like coding tools (as opposed to non coding AI tools) shouldn't rely on AI entirely. We don't need AI to run lint or formatting after each "creative" step. As you said, AI should be trained on LSP and output of those tools.

I've been mainly staying with one company for a long time (8y, 12y, now 4y), and talking to different senior engineers, it's interesting to see how few were exposed to (or even being aware of) a long-term maintenance. Seems like the focus is on creating (esp now with LLMs.)

I'm in the UK and annual tests don't really work. What people do for motorbikes is to ride with illegal exhausts and only swap them for legal before the test. Unless there are random spot checks (like in Austria), annual checks don't change much unfortunately.

Maybe thieves didn't get vaccinated. 😅

For me it's November, not January (because days are getting longer!)

Thank you for the interview! Many other interviews let PB talk about autonomous cars, so it's refreshing to see that there's some love for trains too.

I see, thanks! And good luck and lots of fun with your next book. Also fingers crossed the publisher sees value in including both.