@dajb developers have no idea what they are doing half the time that is why they walk the fine line between being developers and common folk

@dajb

I don't know. If you take the spec of whenwords as a benchmark, it looks much more complex than what you would code in the programming language of your choice. And it has clearly been written by someone who has very good programming skills to avoid the myriad of pitfalls associated with blindly using LLMs.

So, it seems that this doesn't help expert programmers (writing specs is less efficient than writing code) and it does neither help non-programmers (they don't know how to write useful specs).

@dajb same pattern as expressed by @pluralistic

https://mastodon.social/@urlyman/115685951349475474

@dajb @koen_hufkens

When code is cheap, #brand has value

@dajb

"Using an LLM instead of learning to code is a terrible idea. But using an LLM to help you learn to code is great."

https://www.quora.com/Is-ChatGPT-coding-trustable/answer/Ben-Curthoys

@dajb hope you come to the conclusion that " #AI" is irredeemable sooner than later...

@urlyman @pluralistic Oh sorry, I just read the last post in the thread. Yes.

@dajb
what’s the difference?

- increase automation
- reduce human overheads
- hold the remaining humans accountable

@pluralistic

@dajb it's okay to be wrong m8.

• The problem with #AIslop and " #VibeCode|d" bs is that this directly harms others, including myself.

If you what to try something on yourself that's your thing, but don't force me (or anyone else) to suffer from #HypeBasedDevelopment!