Hey everyone, looking for honest feedback on whether I'm solving a real problem or just my own frustration. The Problem I Think Exists: When OpenAI/Anthropic/Google deprecate models (like GPT-3.5 last September), companies with LLMs in production have to migrate everything. From what I've seen: \- Takes 3-6 weeks of engineering time \- Costs $80K-$200K in labor \- Happens every 6-12 months \- Adds zero business value (you're just maintaining existing features) The hard parts aren't just code changes - it's that prompts behave totally differently on new models. What was concise becomes verbose. Sentiment analysis gives different results. Carefully tuned prompts break. What I'm Thinking of Building: Automation platform that: \- Converts prompts automatically (handles behavioral differences) \- Updates API structures \- Side-by-side testing before you switch \- Version tracking \- Goal: 6 weeks → 6 hours My Questions: 1. Is this actually a painful problem? Or am I overestimating it? 2. Would automation help, or is it too custom to each business? 3. What would you pay for this? (trying to understand if it's even viable) 4. Do good solutions already exist that I'm missing? Target Market: Companies spending $50K+/month on LLM APIs with 50+ prompts in production. I've talked to maybe 10 teams, and responses are mixed. Some say "god yes, this is painful" and others say "meh, we just deal with it." Be honest - tell me if this is dumb before I waste 6 months building it 😅 Thanks!