10 days building LLMWatch solo — what I shipped, what flopped, and where I'm stuck

LLMWatch is a solo-built tool that monitors and analyzes the performance, cost, and behavior of large language models in production, designed for AI engineers and developers who need real-time observability across multiple providers. It’s interesting because it tackles the practical pain of debugging and comparing LLM outputs at scale, while the creator transparently shares what worked, what failed, and the unsolved challenges of building such a system alone.