←back to thread

Andrej Karpathy: Software in the era of AI [video]

(www.youtube.com)
1479 points sandslash | 2 comments | 19 Jun 25 00:33 UTC | HN request time: 0.427s | source
Show context
OJFord ◴[20 Jun 25 02:08 UTC] No.44324130[source]▶
I'm not sure about the 1.0/2.0/3.0 classification, but it did lead me to think about LLMs as a programming paradigm: we've had imperative & declarative, procedural & functional languages, maybe we'll come to view deterministic vs. probabilistic (LLMs) similarly.

    def __main__:
        You are a calculator. Given an input expression, you compute the result and print it to stdout, exiting 0.
        Should you be unable to do this, you print an explanation to stderr and exit 1.
(and then, perhaps, a bunch of 'DO NOT express amusement when the result is 5318008', etc.)
replies(10): >>44324398 #>>44324762 #>>44325091 #>>44325404 #>>44325767 #>>44327171 #>>44327549 #>>44328699 #>>44328876 #>>44329436 #
aaron695[dead post] ◴[20 Jun 25 07:14 UTC] No.44325404[source]▶
[flagged]
bgwalter ◴[20 Jun 25 11:27 UTC] No.44326722[source]▶
> It makes no sense at all, it's cuckooland, are you all on crazy pills?

Frequent LLM usage impairs thinking. The LLM has no connection to reality, and it takes over people's minds.

replies(2): >>44326752 #>>44327102 #
boppo1 ◴[20 Jun 25 11:33 UTC] No.44326752[source]▶
>Frequent LLM usage impairs thinking

Is there hard evidence on this?

replies(4): >>44326923 #>>44326929 #>>44327015 #>>44327122 #
1. bgwalter ◴[20 Jun 25 12:08 UTC] No.44326929[source]▶
If you are the type who prefers studies:

https://time.com/7295195/ai-chatgpt-google-learning-school/

Otherwise, read pro-LLM blogs which are mostly rambling nonsense that overpromises while almost no actual LLM written software exists.

You can also see how the few open source developers who jump on the LLM bandwagon now have worse blogging and programming output than they had pre-LLM.

replies(1): >>44328025 #
2. Workaccount2 ◴[20 Jun 25 14:26 UTC] No.44328025[source]▶
I have 7 different 100% LLM written programs in use at my company daily, some going back to GPT-4 and some a recent as gemini 2.5.

Software engineers are so lost in the weeds of sprawling feature pack endless flexibility programs that they have completely lost sight of simple narrow scope programs. I can tell an LLM exactly how we need the program to work (forgoing endless settings and option menus) and tell it exactly what it needs to do (forgoing endless branching possibilities for every conceivable user workflow) and get a lean lightweight program that takes the user from A to B in 3k LOC.

Is the program something that could be sold? No. Would it work for other companies/users? Probably not. Does it replace a massive 1M+ LOC $20/mo software package for that user in our bespoke use case? Yes.