Very sceptical of "Context First: We front-load system understanding before diving into code". The LLM sees the whole input at once, it's a transformer, not a r

Very sceptical of "Context First: We front-load system understanding before diving into code". The LLM sees the whole input at once, it's a transformer, not a recurrent model. Order shouldn't matter in that sense. Ed. I see some people are disagreeing. I wish they explained how they imagine that would work.