tuvistavie

I think having API keys for some third-party services (whatever LLM provider, for example) in a .env file to be able to easily run the app locally is pretty common. Even if they are dev-only API keys, still not great if they leak.

Depending on the country, it can truly be. Some of my relatives have money stuck in Lebanon and there is currently not really any way of getting the money out of the country.

In general, I'd say UK visa process and cost are quite terrible, to say the least. However, for the global talent, the harder part is to get endorsement, after which it should be relatively straightforward. A friend of mine recently got his through the research route and the process seems to have been quite smooth. I think it took about a month or so. For reference, he is finishing a PhD in a top US university and has a good amount of peer reviewed research.

I am not sure I would advocate for this but, from Python 3.8, this is an instance where assignment expressions could allow the first code without duplication.

    first(x * y
          for x in nums
          if (y := 2020 - x) in nums and x != y)

> The difference is that what the doctors put so much time into learning can actually be quite useful for their job.

I'd be curious to hear what a doctor would think about that. My intuition would be that, as for almost any exam, a lot of time is spent on things that will not be useful for any other purpose than the exam itself.

Runtime safety and compile time safety are not the same thing. Compile time safety is not a silver bullet but it does protect against quite a few classes of errors.

I think the only possible answer to how much math do you need is that it depends on what you work on. There is a lot of stuff that can be done only using very basic math, that's for sure.

I would tend to agree with the article that math can be quite useful as a general mental model and can often help to find cleaner solutions to some problems.

There are also a lot of applications that require some, often quite specific, knowledge of math: machine learning, cryptography, game engines, to name a few.

As much as I agree with you on all the points, in my case I kind of have to stick with what my research group is using for now.

I'm in the middle of my PhD and switched from printing papers to reading them on my iPad roughly two years ago.

There are some things I miss from paper but overall I found the pros to overweight the cons.

I haven't found that flipping back and forth on iPad is that horrible, to be honest.

Not sure how helpful this will be but I'll share what I've been doing for now. I use the following apps:

* Mendeley (to organize papers)

* PDF Expert (to annotate PDFs)

* GoodNotes (mostly when working out the maths)

My usual workflow is:

* Read through the paper

* Annotate in the paper using Apple pencil as I read through

* Figure out the maths on the iPad when needed

* When I get back to a computer, upload the annotated file to Mendeley and type summary notes in Mendeley

A few things that I like/dislike about iPad when compared to paper.

+ Search for information on the web while reading paper more easily

+ Check notes/annotations quickly from my computer

+ Share notes easily

+ Search notes easily

+ Clean desk =D

- More context switching needed when I need to scramble something

- Mendeley misses some basic features on iOS (e.g. attach PDF to existing paper) so need to context switch with computer at some point after reading the paper

I would say that for 90% of the papers I go through, where I don't dive that deep in the paper, the experience is just as good on iPad. For the 10% of the papers I read where I go in-depth, redo proofs, etc, it's a little more tedious. While it's for sure not perfect, given the above pros, I can live with the cons.

At the end of the day I think it's just a matter of cost. If it's cheaper to let a software fail and simply fix it when it does, there will not be much interest into proving it to be correct. If a failure will likely be vastly more expensive than proving correctness, then proving correctness will make sense. There are simply way less instances of the second category. I think one of the few "new" instances, which might be worth mentioning is smart contract programming, where failures can cause millions of loss and proving correctness is not too expensive.

I am pretty sure we reached this point quite a few years ago. From what I know, most people learning web programming nowadays start with Rails, Express JS or whatever framework of the day they have been recommended, and use the integrated HTTP server for development, I would assume, often without even really worrying about what is actually happening. CGI scripts are definitely not on the list of things to learn there.