6 ms·
It’s getting better on Claude specifically now that training cutoffs are starting to move beyond the end of 2023. OpenAI models are a little less likely to tri
by F7F7F7 11mo ago
It’s getting better on Claude specifically now that training cutoffs are starting to move beyond the end of 2023.
OpenAI models are a little less likely to trip up as long as you keep reminding it that it needs to account for v0.23+
- throwaway77385 10mo agoI've found Gemini 2.5 pro to handle it ok. It forgets small things like using the deprecated model instead of baseModel and things like that, but nothing dramatic.