Across teams and organizations what I’ve seen this year is that the “best” developers haven’t looked. The proverbial rockstars with the most domain knowledge ha

Across teams and organizations what I’ve seen this year is that the “best” developers haven’t looked. The proverbial rockstars with the most domain knowledge have never gotten around to AI and LLMs at all. This is compounded by adherence to misguided corporate policies that broadly prohibit use of LLMs but are meant to only be about putting trade secrets into the cloud, not distinguishing between cloud vs locally run language models. Comfortable people would never challenge this policy with critical thinking, and it requires special interest to look at locally run language models, even just to choose which one to run. Many developers have not advocated for more RAM on their company issued laptop to run better LLMs. and I haven't seen any internal language model that the company is running in their intranet. But it would be cool if there was a huggingface-style catalogue and server farm companies could have and let their employees choose models to prompt, always having the latest models to load.