@baldur The only case I've seen in LLM used for coding in a way that makes sense was in a way where it was clear it was still and incredible amount of work and was used not to churn stuff out, but to do things that wouldn't be feasible otherwise.

They describe LLM in the context of coding as dangerous power tools that should be used with care and consider pushing it on novices as bad.

Personally I think the dynamite is a better analogy than power tools (think early dynamite that accidentally exploded easily).

I think when the dust settles maybe we might see some specific instances of these tools that are ethically trained and run, and that have very specific uses cases that actually work. By that point I think those should be very obvious though, and until they are I'll remain extremely sceptical.

But like always in tech (and like you say in your book): for most people there's rarely a big advantage to be early adopters.

EDIT: Outright formulated one of my sentences in a way I completely disagreed with 🙈

@baldur I feel a strong pressure to use LLMs. The reasoning is simple.

I want a career in tech. I have wanted this for a long time, for years. The skillset I have is entirely oriented toward that.

Many companies have spend an enormous amount of money now to put out the message that if that is what you want, then __you have no choice__ but to embrace LLMs.

They are communicating as forcefully as they can: LLMs or healthcare, LLMs or a home, LLMs or a future.

Making an ethical living was hard enough before LLMs. It is harder now. I do not know how long I can hold out.

@baldur
Far as i can tell, the productivity case against LLMs for work is as simple as empirical review:

LLMs do not provide productivity gains.