> Is there some expectation that these things won't improve?
Sure. But the expectation is quantitative improvement - qualitative improvement has not happened, and is unlikely to happen without major research breakthroughs.
LLMs are useful. They still need a lot of supervision & hand holding, and they'll continue to for a long while
And no, it's not "ego speaking". It's long experience. There is fundamentally no reason to believe LLMs will take a leap to "works reliably in subtle circumstances, and will elicit requirements as necessary". (Sure, if you think SWE work is typing keys and make some code, any code, appear, then LLM are a threat)
Sure. But the expectation is quantitative improvement - qualitative improvement has not happened, and is unlikely to happen without major research breakthroughs.
LLMs are useful. They still need a lot of supervision & hand holding, and they'll continue to for a long while
And no, it's not "ego speaking". It's long experience. There is fundamentally no reason to believe LLMs will take a leap to "works reliably in subtle circumstances, and will elicit requirements as necessary". (Sure, if you think SWE work is typing keys and make some code, any code, appear, then LLM are a threat)