> While these checks could also be handled by unit tests, most scientists generally just end up with their own weird set of ad hoc test outputs and print statements. It’s ugly, and not infallible, but it tends to work well given the intensive nature of our result-testing behaviour and community cross-checking.
The sense in which this statement is plainly wrong is actually addressed in the original article:
> Regressions like that are common when working on a complex piece of software, which is why industrial software-engineering teams write automated regression tests.
This is the whole point. You can only run your "ad hoc" print-statement-based tests once off, which is why such "tests" are useless for finding regressions.
There's a bias on HN to compare working-from-home with the commuter experience _in the Bay Area_, and conclude working from home is better. For many of us in places like New York, living in the city and going to the office is much more preferable lifestyle than working from home in a random place.
100% percent my experience too. In the last 3 months I've been on-site at 3 of the FAANG companies. Not a single interviewer brought up the large personal projects I've done and which are given decent space on my resume; my weeks of leetcoding was definitely far far far more beneficial for getting a job.
I'm on a H-1B visa, which sounds like the same visa your co-worker is on.
H-1Bs are permitted to change jobs, and if they are laid off from their current job they have 60 days to find new employment before having to leave the United States. Even if they have to leave, they can continue to job hunt remotely (really hard, I know) and if they get an offer they can return to the US on the same H-1B (no lottery). In NYC at least, medium to large employers are very happy to process the H-1B transfer paperwork because good engineers are in such high demand.
If your co-worker is as good as you claim he is, he should start looking for a new position. It's not as hard as many people think it is on this status. The main thing is to get beyond the mentality that you're beholden to your original sponsor - you aren't.
> While these checks could also be handled by unit tests, most scientists generally just end up with their own weird set of ad hoc test outputs and print statements. It’s ugly, and not infallible, but it tends to work well given the intensive nature of our result-testing behaviour and community cross-checking.
The sense in which this statement is plainly wrong is actually addressed in the original article:
> Regressions like that are common when working on a complex piece of software, which is why industrial software-engineering teams write automated regression tests.
This is the whole point. You can only run your "ad hoc" print-statement-based tests once off, which is why such "tests" are useless for finding regressions.