I cannot stress how important it is to engage with Rafiki and correct him if he ever says anything off

He is right 80% of the time. The other 19% must be fine tuned

Eventually, he’ll be 99% correct (which is where most 4.0 and 5.0 models like ChatGPT are today)

The hallucinations are always bad in the early days but as fine tuning takes place, it gets significantly better each day

80% is a great start. Better than most 1.0 AIs if you look around

Once I am happy with Rafiki’s responses 90% of the time, we have a pretty sick UI update in the pipeline for INLEO

This will include direct-interfacing, searching, organizing, etc. a plethora of features

Continue to fine-tune so we can get to 1.50 faster. When we’re at that place, I’ll release the new features that I think will blow your mind even more-so than Rafiki’s mere existence

Imagine something like Grok on X but a 10x better interface