They are currently experimenting with replacing engineers with LLMs, so it would at least be interesting to try the flipped experiment.

It seems like, at least for tasks like distilling the work done at lower levels and passing it up the management tree, LLMs should be good at that. Maybe they could take out some levels at least.

And they can be programmed to be 100% selfless and on-mission: no turf battles or other internal politics. Everybody’s job involves office politics of course, but my gut says management has a lot more of it than engineering, so zeroing that out could be more significant.

Aren't LLM's mostly overconfident? If they are independent entities/agents, they may be tempted to try to sabotage the less performing teams to take over their projects - just like in human world.

The models are trained on data that encodes hundreds or thousands of years of turf battles and office politics. The model is shaped by the script of ‘Succession’, the history of the English monarchy and the Chinese Communist Party, and the biographies of Steve Jobs and Alexander Hamilton…

The models don’t generate what’s right and good: they generate what people are likely to generate.