small_model

14 ms·

1. 241.▲
   by small_model 8mo ago

   Would two great developers produce the same output given a spec, probably not. The spec and validation would be part of the solution. The code can be different each time.
2. 242.▲
   by small_model 8mo ago

   Indeed, his name is Claude
3. 243.▲
   by small_model 8mo ago

   So you know every line of code in every framework/library/sdk you use! wow
4. 244.▲
   by small_model 8mo ago

   We will see, I still think we will need engineers who understand fundamentals, and yes we need to teach new devs these like we teach them how assembly, computer architecture works, even though we likely don't build our own compilers or
5. 245.▲
   by small_model 8mo ago

   I mean the code would be intermediate and not really kept around. Code could be any language llm would choose best depending on requirements.
6. 246.▲
   by small_model 8mo ago

   I disagree, and even more so in the future. Programming languages will become a relic of the past. I think LLMs will just create binary from spec/tests/
7. 247.▲
   by small_model 8mo ago

   Coding session these days is someone typing into Claude Code and waiting. I think this would have been a decent idea a few years ago. Humans typing code into an editor is going to be as rare as debugging assembly.
8. 248.▲
   by small_model 8mo ago

   So using weedkiller isn't gardening to these people?
9. 249.▲
   by small_model 8mo ago

   Same as assembly programmers felt when C came along I guess
10. 250.▲
    by small_model 8mo ago

    The proper analogy would be you can now remove all weeds with the swipe of your hand and cut all your hedges with another swipe, you still are gardening you can do it quicker and therefore explore different possibilities.
11. 251.▲
    by small_model 8mo ago

    Who will buy this when the X roadster is released that is able to hover etc from SpaceX tech? Legacy auto is toast they move too slow.
12. 252.▲
    by small_model 8mo ago

    Every repo usually has at least one local copy somewhere, worst would be few old repos disappear.
13. 253.▲
    by small_model 8mo ago

    Why would you think he is not ok? Just linking to a GitHub repo he maintains doesn't really tell the story.
14. 254.▲
    by small_model 8mo ago

    We would still need the language to be human readable, but it could be very dense. They could build the ultimate std lib, that goes directly to kernels, so a call like spawn is all the tokens it needs to start a co routine for example.
15. 255.▲
    by small_model 8mo ago

    Claude is only a few years old so we should compare it to a 3 year old human's C compiler
16. 256.▲
    by small_model 8mo ago

    How about we get the LLM's to collaborate and design a perfect programming language for LLM coding, it would be terse (less tokens) easy for pattern searches etc and very fast to build, iterate over.
17. 257.▲
    by small_model 8mo ago

    Ok thanks, hopefully, its annoying to lose or have context compacted in the middle of a large coding session
18. 258.▲
    by small_model 8mo ago

    I have the max subscription wondering if this gives access to the new 1M context, or is it just the API that gets it?