We have some interesting evidence from AI agents. People used to make the plan with a stronger and more expensive model, and then use a cheaper model to do the "grunt work" of implementing it.
But it turns out it's better to do it the other way around. A good plan is easy, implementing it is not.
> plan with a stronger and more expensive model, and then use a cheaper model to do the "grunt work
Doesn't this sound like the model that led to bloated poor software produced by code monkey sweatshops where an architect fobbed off implementation work to grunts?
It always brings me back to the anecdote that QQ had 6000+ devs and WeChat was built by a tiny core of elite engineers (by imitating WhatsApp)
Really? I'd love to read some write ups on this because I'm still doing it the old way. I was going to say that corporate experience runs counter to this but maybe not; you think the architect is more important than the developer, but actually there have been shifts over time to recognise ICs more and keep them in their roles while managers are more generic. Interesting food for thought!
But it turns out it's better to do it the other way around. A good plan is easy, implementing it is not.