lysecret

21 ms·

1. 331.▲
   by lysecret 4y ago

   We always use dataclasses with frozen=True to make them immutable.
2. 332.▲
   by lysecret 4y ago

   Pydantic is much much much slower than normal dataclasses. However, they have to be because they do validation and dataclasses do not. So, I use normal dataclasses for internal constructs, which need to be fast and pydantic for anything tha
3. 333.▲
   by lysecret 4y ago

   I love notebooks for me there is no better way to dive into a new dataset or badly documented api and just try and error your way through. Keeping whatever messy state you want. I always see them as “throw away” anything that is useful gets
4. 334.▲
   Docker Beta Feature: Rosetta for Linux on Intel-Based Images on Apple Silicon(docs.docker.com)

   2 points by lysecret 4y ago|2 comments
5. 335.▲
   by lysecret 4y ago

   I randomly discovered this new beta feature. Because of some very specific libraries we have to run x86 images. This new update has increased build times and overall code execution by a factor of 10.
6. 336.▲
   by lysecret 4y ago

   Can confirm been working with F# for 2 months and I really miss the strong type system, DU and match statements when I go back to Python.
7. 337.▲
   Ask HN: Aren’t high Egress fees clearly anti competitive?

   17 points by lysecret 4y ago|9 comments
8. 338.▲
   by lysecret 4y ago

   Absolutely love the first one. His explanation of SVD had a big impact on me and my dad is now teaching it in his math class haha
9. 339.▲
   by lysecret 4y ago

   I am not a big fan of his work anymore but the original Personality Lectures (2017) of Jordan Peterson are really great. https://www.youtube.com/watch?v=kYYJlNbV1OM&t=1s
10. 340.▲
    by lysecret 4y ago

    Funny to me ETL was always about getting data to a format where you can then use SQL to run whatever on. ETL means a lot of very different things to different people.
11. 341.▲
    by lysecret 4y ago

    My opinion. Always if in any way possible pick a semantic key. There is usually something defining the thing you are working on. If there isnt work on your normalisation. Main benefits to this: Avoids accidental duplication (happens so much
12. 342.▲
    by lysecret 4y ago

    100% Signals an Threads. Super technical extremely well presented. Some more fluffy ones: Hard fork. Pivot. All in.
13. 343.▲
    by lysecret 4y ago

    Wait what this is absolutely brilliant. Actually insane it works so well using a stack of iPhones as an ocr server. My deepest respect.
14. 344.▲
    by lysecret 4y ago

    Well the minimum cost of Alloydb currently is 600 dollar a month. We don’t want to spend that amount on dev and staging environments.
15. 345.▲
    Ask HN: What's Your Experience with AlloyDB?

    2 points by lysecret 4y ago|2 comments
16. 346.▲
    by lysecret 4y ago

    Absolutely love ChatGPT. I am currently learning F# and it has been a grandiose help.
17. 347.▲
    by lysecret 4y ago

    Imo as useless before as after. Software engineering is life long learning. 5 years in school is better spent solving real problems anyways. All these Ai advancements help you solve more problems. Just like higher level languages or your fa
18. 348.▲
    by lysecret 4y ago

    Also much prefer cloud schedule over AWS version.
19. 349.▲
    by lysecret 4y ago

    Big fan of retool too. I also like to throw FastApi in front of Postgres on lambdas or cloud runs.
20. 350.▲
    by lysecret 4y ago

    I know exactly what you feel like and I used to be exactly like you. Even down to preferring R to Python. All I can say is: Bite the bullet. Learn Python, forget about R. R is nice and all but there is nothing that couldn't also be don
21. 351.▲
    by lysecret 4y ago

    Just put it as an environment variable.
22. 352.▲
    by lysecret 4y ago

    Their podcast "the joy of why" is still top notch.
23. 353.▲
    by lysecret 4y ago

    Hmmm. Not sure why they use M3 data when there is already M4 where a deep learning model won. I know because I reimplemented it as a toy version in python here: https://github.com/leanderloew/ES-RNN-Pytorch It was actu
24. 354.▲
    by lysecret 4y ago

    Interesting article, I immediately had to think about terraform. The problem I'm having with specialized languages (even with SQL) is that it always creates additional interfaces and almost always creates ugly string formatting if you
25. 355.▲
    by lysecret 4y ago

    Wait will the forced liquidation crush a large part of crypto ?
26. 356.▲
    by lysecret 4y ago

    Nothing is "just" one thing. But the there is a reinforcing effect which definitely makes it worse. You can check out "animal spirits" and Keynes if you want to go down a rabbit hole ;)
27. 357.▲
    by lysecret 4y ago

    One thing to always remember is that companies (especially startups) have strong incentives to always have a lot of job openings on their websites. Because it looks like the company is doing great. Also, there is no legal obligation to actu
28. 358.▲
    by lysecret 4y ago

    Might be interesting to look at finance / trading related things since they have a pretty strict trading day (I think).
29. 359.▲
    Will Rust Take over Data Engineering?(airbyte.com)

    4 points by lysecret 4y ago|0 comments
30. 360.▲
    by lysecret 4y ago

    Companies job pages are very vague estimates of actual offering, especially at startups were everyone is incentivized to post more than existing to look good.