Seems excessive.

Directors of small companies are the same, they're just not wealthy enough that they could do any harm.

Most smart juniors have no problem with learning. Perceptual exposure and deliberate practice works almost mechanically. However, if someone can't tell you what examples you should be exposed to, you'll learn crap.

My first thought was, well agents seem nice, but I think, AI workflows are a better bet. However, I don't really understood AI or agents in depth and felt like I was just "doing things the old way" and removing flexibility from agents was a ridiculous idea.

After some research I got the impression that I was right. A well defined workflow and scope is just what's needed for AI. It's cheaper and more consistent. It probably even makes the whole thing run well with non-SOTA models.

GPT and Claude would work much better than Gemini, even if the direct feedback was sparse or diffuse.

However, the moment I gave Gemini a fast testing framework that gave it instant feedback, it would mill through all kind of problems.

Claude and GPT are seniors.

Gemini is a very motivated mid level.

Anthropic even says, that an agent based solution should only be your last resort and that most problems are well served with a one-shot.

https://www.anthropic.com/engineering/building-effective-age...

You can use it with just European models if you want.

No, it aimed to solve the "out specs are bad and we need to iterate faster" problem.

"a massive lie exposed by LLMs"

No. LLMs add no insight about the problem and they expose nothing. They just help to engage this well-known problem with another tool.

I liked Ra, but I liked Fine Structures more.

Later I read the first version of the book and it was okay, but the vibes were a bit lost.

The new version of the book I didn't even finish.