From my hikes this weekends!

Yeah this is definitely what I love about typescript. Done right, it can provide an awesome developer experience.

Attempting to go 0-100? I found no success with that. It’s the most successful when I’ve written code and I need help: transforming, refactoring and writing tests. It does better with having code context, but not amazing.

answer: cursor's rules for AI - though unless you have yolo commands enabled, you'll have to manually hit the run test commands.

I’m wondering if I just need an extension that runs the tests and surfaces the errors in the editor. It seems to be able to detect and fix lint errors without issue!

I would not say struggle, but it doesn’t seem to be using the context of previous tests well.

For example, I have a Go test that requires passing in a typed nil value instead of “nil” into some methods. It for some reason never picks that up on the first go in generating tests, always uses nil.

The transformer use cases are the best, probably the most common use case for me with regard to LLM generated code. That and generating tests. But tests still need so much wrangling, even if I have context of previously written tests.

Only lately have I felt that this is probably very true. Hasn’t always been the case, I never used to think it mattered much before. But maybe the boring is now so ubiquitous!

This is what I assumed is happening (excluding the transform flag). For node-only code, shouldn’t this be better? No transpiling necessary. You write modern JS with types. Unless I am misunderstanding how this works.

This is pretty cool - love it.

Precious time!

tailwindcss.com/blog/tailwin...

With tailwinds making it easier to style with “just css”, this make shadow dom style encapsulation less compelling to me. Less to wrangle with, and probably easier styling DX. I’ve yet to experiment though.

It’s amazing that things like this cause so much commotion. Yet things like easy foot guns people are less noisy about (e.g crazy reactive hell in previous svelte versions had. Or react hooks…). Build large code bases with various exp levels, then these seem like such nonissues.

I think you guys got the price point right on. Loving the ad free experience.

So who, in your opinion, does this well?

Splitting out products makes sense to me. But something like forcing an open access API, that seems off - I’m not sure why. Something about forcing companies to grant access to user generated data feels off.

Had that same feeling when building an email viewer component! Shadow dom encapsulation made style isolation a lot simpler. Of course, it broke other things (like custom fonts :))

Is there a market for AI services that upgrade these apps to modern versions? I feel like the cost to make such an upgrade might be significantly reduced. Cheaper than a specialized dev!