🎯 Getting good results from LLMs

Building with LLMs can take many different paths with widely varying results. At one extreme, AI slop; at the other, work better than anything we’ve done before. What is the difference?

Many are obsessed with better models and tools, looking for the silver bullet that will produce exceptional results with no mistakes and little effort on our part. Most of our industry news about LLMs is focused on this. However, I don’t think better LLMs and tools are the key. They are all pretty good now. Exceptional, accurate, safe, and reliable results come from the platform, architecture, and abstractions you build around the use of LLMs. You need to own and understand this, and allow the LLM to work inside this framework, like you would any team of engineers. Anything else becomes uncontrolled and unmaintainable.