Thanks I do the prompts in langfuse and then test multiple agents vs the prompts to see which ones perform better; this is for very specific use case which is an AI strategy bot - whose aim is to provide non technical ai strategies for enterprises across multical verticals that can be used by middle managers (non technical) as a basis to start deploying AI in their businesses. For general research, I tend to use chatgpt/claude/perplexity. And because I work with big tech platform commerce infra, tend to be vertical (eg gemini for g, meta.ai for, meta) because i assume their training data includes a lot of hard to find developer documentation. I'm much more down the maturity curve than you which makes me laugh because I describe myself as '0 to 1' product person in my various projects.
Thanks I do the prompts in langfuse and then test multiple agents vs the prompts to see which ones perform better; this is for very specific use case which is an AI strategy bot - whose aim is to provide non technical ai strategies for enterprises across multical verticals that can be used by middle managers (non technical) as a basis to start deploying AI in their businesses. For general research, I tend to use chatgpt/claude/perplexity. And because I work with big tech platform commerce infra, tend to be vertical (eg gemini for g, meta.ai for, meta) because i assume their training data includes a lot of hard to find developer documentation. I'm much more down the maturity curve than you which makes me laugh because I describe myself as '0 to 1' product person in my various projects.