@BCSkiingRMNP wasnt expecting anyone to take me seriously, but cool to know someone on here deep in this stuff. also a software guy but not in the AI space at all.
what are your thoughts on DeepSeek? just heard of them recently, being billed as “what OpenAI promised to be”. their benchmarks look good, but I haven’t looked into the actual code yet.
also, where do you see levels of abstraction stabilizing? i agree, a “fine tuned model” can go a pretty long way, but I feel like a lot of the actual work/coding that goes into tweaking data sets, adjusting feedback loops and weights, etc, is being abstracted into dumb user friendly things like “Personalities”. in other words, as we find and train models to be better at doing specific jobs, do we all just become prompt engineers?