Hacker Newsnew | past | comments | ask | show | jobs | submit | ndand's commentslogin

There is a way to watch the video anyways, on YT with just 2 clicks.


> Valued = useful + hard to find.

I agree, but I think it slightly differently, Valued = useful + hard to replace.

How difficult is to be replaced by some other human or a machine.


I'll add one more to this, "Pleasant personality". If you have are difficult to be with, you will be miserable and so will others around you, leaving you with very little or no allies. So new formula:

Valued = Useful + hard to find + easy to work with


"Easy to work with" is a wide requirement. It's harder to find.

But we can compare how far "easy to work with" goes in a more balanced scenario:

You're hiring for a critical rare skill. _Only two_ candidates were seriously considered, and you hired them both.

One of them delivers for the skill you hired (useful), but is not easy to work with.

One of them does not deliver for the skill you hired (useless), but is easy to work with.

So, it doesn't matter what your definition of "easy to work with" is in this scenario. It justs matters that whatever it is, it's not related to usefulness (so you can't get away with "but being useful makes you easy to work with").

Who would you lay off if your bills were depending on those guys?


I understand it differently,

LLMs predict distributions, not specific tokens. Then an algorithm, like beam search, is used to select the tokens.

So, the LLM predicts somethings like, 1. ["a", "an", ...] 2. ["astronomer", "cosmologist", ...],

where "an astronomer" is selected as the most likely result.


Just to be clear, the probability for "An" is high, just based on the prefix. You don't need to do beam search.


They almost certainly only do greedy sampling. Beam search would be a lot more expensive; also I'm personally skeptical about using a complicated search algorithm for inference when the model was trained for a simple one, but maybe it's fine?


You can capitalize a string in Python using functional style

  ' '.join(map(str.capitalize, string.split(' ')))
which is similar to the example in Ruby, except the operations are written in reverse order.


Google has a few courses about technical writing for engineers:

https://developers.google.com/tech-writing


Steve Yegge mentions the book in the "Compilers, why learn them?" episode,

https://www.youtube.com/watch?v=o87pOVK8khc&t=1939s

"you're not done actually when you learn how compilers work you also need to learn how linkers and loaders work and you need to learn how to operating systems work before all of the magic is gone. So you really want to learn compilers and operating systems and then get this book that I have called Linkers and Loaders it's like the only book on linkers and loaders. I should have brought it and uh and it's really good"


> I believe there's now also a VS code plugin for the same behavior, but I don't know what it's called.

It's called "sticky scroll". It's a relatively new feature of VS Code.


Not for the rest of their lives. 50 people can live on $100k/year for 20 years.


if you assume typical stock market returns, ~$2M is enough to retire on and live on $100k a year of interest gained alone.


Assuming you are using recent interest rates. If you mix in recent inflation rates too it doesn't work out in inflation-adjusted dollars like that, though hopefully a lot of that is transitory from covid (or maybe we have something worse than covid in the future).


"Roberta and I will be on Twitch tomorrow to talk about a new game we're working on. Our interview will be at 3pm Pacific time."

https://twitter.com/caboken/status/1506390567197896705


My attempt to make a practical keyboard-driven cursor app for Windows:

https://github.com/ndandoulakis/SlickCursor


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: