This isn’t the first prompt or query it’s done this one, or file for that matter.
It seems to spend a tremendous amount of its own brewing time on matching whitespace, which is the least relevant thing when I’m looking for help from it.
I like to watch it’s thought logs, and I can see it strike upon a solve relatively quickly, but then about 90% of its bake time is just presentation. Unacceptable imo.
I’d like it to minimize it’s focus on whitespace, even when I add such an instruction to my query, it’s only brought up once the AI has already spun the block once or twice on whitespace matching. And even then, it’ll go back to whitespace matching one step later. Wild.
Expected behavior
I’d like it to minimize it’s focus on whitespace from a baseline perspective. Results now are better than pretty after a while. Especially when our daily limit is at risk.
Hi, we’re aware of this issue and are currently testing a fix. I’ve overridden your account with the patched version. Please let us know if you run into the issue again. Thanks!
Honestly, I feel like if the assistant should ignore tabs, or have some sort of robust linear tool to determine the indent level of a given tab, rather than relying on its own parsing.
Hi @TheFaeTwins, I was able to reproduce the issue and have switched your account to a different setup. Please let us know if you’re still seeing the problem.
It looks like the tools are doing right thing, but the model may be stripping out tabs and empty lines.
Okay, I’m not sure if I can take this off solved, but while it is remarkably better, it’s definitely tripping on multi_edit requiring exact whitespace matches:
