The “Liberation Day” Tariffs.

Well, gifts.

No tax on donations.

Sure I can sell you some Blockstream stock.

https://github.com/lightningdevkit/rust-lightning/pull/4297 I believe vinchenzo is hoping to ship a one-method library with this soon to give people something they can integrate for validation as a first step.

It seems to depend a ton on what kind of project it is. Complicated nuanced state machine like lightning? Not so great. More straightforward things, or UI? Really really good. It’s often faster and better at the strict “code generation” step but that was always a relatively small (esp cognitive) part of software engineering/being a developer.

It helps, but it’s far from perfect. Especially with complicated state machines it can really struggle with things that aren’t wildly complicated (cause they lack formal reasoning).

Nah, even with good coverage (though we aren’t great now) it makes really poor decisions for new API designs sometimes. Just like with English it’s *really* good at writing something that looks very correct. With code that’s often actually correct and pretty decent (if occasionally verbose), but certainly not always, and definitely not always the best or most maintainable design. Sure, you can “prompt better” by giving it the specific design you want, but usually the only way to find the right design is by exploring options. LLMs are useful to speed that process up, of course, but claiming they’re replacing that process today is absurd.

Ultimately review is generally a much more difficult skill than code authorship, so there’s a tradeoff for subtle changes.

Depends a ton on what for. For small applications, scripts, tests, self-contained things, or repetitive tasks (think refactoring), it’s basically 100% AI+review (often tests you have to tweak and carefully check cause it’ll write tests that pass but don’t actually test anything with it being obvious at all!). For high-level tasks in large codebases, especially starting to design a new API or feature, it’s somewhere between 0 (doing it yourself to explore an approach while being more directly exposed to the changes) or using them to explore but then rewriting chunks after you’ve explored different options.