OCR is one we've investigated but it's tricky to find the right API that is compatible with our local-first values. One customer tipped us off that the creators of Nebo have an interesting one. So we'll look into that.
Sorbet is very useful, but the ergonomics suck. It’s fucking difficult to write rspec tests. The performance overhead of writing sorbet on rails in a big codebase is so much that we have turned it off. The pre-interpreter type checking is somewhat useful.
The alpha releases are also a big concern. We are stuck on a 300 commit (release) old build and can never upgrade safely.
We have also never been able to get the VSCode extensions to run.
Thanks for Sorbet, but I’d suggest people outside Stripe to look elsewhere.
Don’t want to be cynical (and appreciate your more constructive feedback), but i’m surprised by the kind of stuff that makes up to the front page these days.
I tend to agree. I don't expect every project to have a landing page with flat trendy graphics giving you the bullet points like a start up, but just a few paragraphs/link to a white paper would go such a long way.
Isn’t this just GPT-3 under the hood? Other similar things do exist (eg: copy.ai).
Not sure whether there’s much of a difference between all these GPT-3 powered services when all that distinguishes you from competition is some (slick) UI and the 500-1000 extra words of “training” you give to GPT-3.
Gwern [1], who has spent quite some time with GPT-3 and previous models, seems to think that coming up with the right 500-1000 words can be a subtle business.
Now, what's the prompt that will get GPT-3 to generate good prompts for us?
https://arxiv.org/abs/2102.07350 already calls it "metaprompt" :) I gave it a quick stab a while ago but I think prompt programming is too new, and you can't easily cram demonstrations into an existing prompt, for it to really work well. It's more promising to train models on examples of tasks from instructions, or work on directly optimizing prompts for a goal (https://arxiv.org/abs/2101.00190) - it's a differentiable model and a whitebox, so use that power!
I would ask people to stay away from Starsky. Terrible recruiting practices -- I met some of their senior staff and they assured me, well even guaranteed that someone will get back, but no one did.
As job-seekers, we should start maintaining a black list of firms that treat prospective employees badly, and urge others to not apply.
Can you email me at daniel@starskyrobotics.com? I'm sorry to hear you had a bad experience - that does not match how we aim to recruit and no one should get lost in our process. I personally hate companies ghosting people - we don't do that.
It feels like you’re bending over backwards to mollify this person only because you got called out publicly and you’re getting downvoted because of it.
I don’t even know how emailing you about it is supposed to help. If GP has been ghosted by you years ago, in what way does it help them to contact you now? All it does is provide an easy way for you to improve your hiring process while giving the other person nothing in return. The dynamic is similar to the unpaid internships we’re railing about in this thread.
I expressed my concerns about not hearing back from employers to Kartik -- I applied only because he reassured me that every application "in their system" (sic, iirc) gets a response.
"We don't do that", hah. If a founder wants to lie to prospective hires, that too when they are meeting them for the first time, it suggests that something is really rotten at the company. If you don't treat people nicely, well, there's a tonne of other firms in the area that are hiring...
What's your background, and what position was the company looking to fill?
I was accused of this same thing on a "Who is Hiring" thread a while back. I posted about looking specifically for senior front end engineers. What I got instead was close to 50 emails from boot camp grads, many of them data science.
If candidates feel the need to disregard what I say, why exactly do they deserve my time in return?
Ok. Throwing someone's words back at them with "hah" is being a jerk. Claiming someone "wants to lie" when there are plenty of other obvious explanations is being a jerk. So is venting your frustration on some dude just because he happens to work for the company, especially when he was trying to be nice and you prefer to assume bad faith. So is overposting about this instead of saying your piece once and letting it go. And actually, acting like your bad experience generalizes universally and justifies your being vengeful, is also kind of being a jerk.
Interesting rationalization. Thanks for putting it into words.
I agree that my tone was could have been better.
I don't buy the rest of your arguments, and would not comment on them for the sake of being misinterpreted again -- following the Golden rule of not feeding the trolls.
Just finished reading DDIA, can't recommend it enough! I learned a lot of new info in every single chapter even for topics I thought I had a firm grasp on. Great job Martin Kleppmann!
I know this sounds like a cliche at this point, but volume 3 of Donald Knuth's "The Art of Computer Programming" goes into more depth on the theory that underpins these algorithms than anything else I've ever seen/read/heard of (in fairness, I haven't read OP's suggestion yet, though).
This is one of the best technology books I've ever read. I spent a few years diving into big data architecture. I thought I had a reasonably good handle on it, then I read this book. So many insights.
