Interesting.
Is there any way you can chart some data somehow that shows the improvement along side when LLMs got good?
Thing is LLMs got sure usable 'bout Jan this year, but I think they were capab...
I have no hard data but at @WalletScrutiny we are working on reproducibility. It's our bread and butter and our engineers are heavily using LLMs all the time to reproduce. Without LLM, we used to tell wallet providers "it failed to reproduce" and had very limited resources to dig really deeper and suggest improvements. Now it's just some prompts to an LLM more to get a well researched issue opened with the providers.