Reaction
Haiku 5.5 versus DeepSeek on Lindy's evals
The poster says their prompts may be optimized for DeepSeek after long use; they're running GEPA on Haiku.
TLDR
A post shares a Haiku 5.5-versus-DeepSeek comparison on Lindy's evals. In a follow-up, the poster cautions that long use of DeepSeek may have optimized their prompts for it, though they consider that unlikely because they say DeepSeek is mostly distilled from Claude. They say they're running GEPA on Haiku with their prompts and will report back.
Combined views
2.1K
2 Sources, first seen ago
