#Honeslty a bit annoying haha, its still
drifting carbon
Which model is this?
Ah I see anthropic but is it 3.5 or 3.7?
(and hopefully not Haiku haha)
rose blaze
Ah I see anthropic but is it 3.5 or 3.7?
Its 3.5 sonnet i think, whatever is the default, should I try 3.7?
drifting carbon
worth a go, but might not be a silver bullet. 3.5 is the default yeah. claude-3-7-sonnet-latest for 3.7
do you have code or a trace anywhere i can see?
rose blaze
Yeah for sure here’s a good one https://v3.dagger.cloud/levlaz/traces/9c9d334943688922a5f9e23881b680ae
Every once in a while it does the right thing, I also tried Gemini with mixed results
drifting carbon
ive been impressed with gemini-2.5-pro-exp-03-25 a few times
also try out -i if you find it stopping to ask a question all the time - that'll let you interject
unborn burrow
ive been impressed with `gemini-2.5-pro-exp-03-25` a few times
the sentiment I got from X from the last few weeks is:
- 3.5 is better than 3.7 in some things, mostly code generation
- gemini 2.5 pro beats 3.5 and 3.7 if you're very explicit
should actually link back to the posts/benchmarks somehow
re: 3.5 vs 3.7
3.7 tends to suggest/do more than what it's being asked for
I felt this using cursor myself too