Google OKF and ICM
So, for the last week or so I've been digging into the OKF from Google. This past weekend Jake brought it up in his Afternoon Tea session.
It filled in the one thing that I wasn't grasping on my own.
Its not a alternative or a replacement for ICM, its an additional layer that drops right in.
Much of the OKF is built like ICM stuff, which makes sense. The big difference is that its designed about retrieving data and then feeding that into the Folder Agent.
Think of it more like an Index at the back of a book that tells the folder agents, where data is and how to get it. It can be local or it can be out on the internet.
Jake used an example of a topic and looking for it on Stack Overflow at this weekend's Afternoon Tea. A "traditional" ICM agent/specialist looking for that topic on Stack Overflow, would have to search the whole thing. With the OKF layer in place, the specialist sees there is a link with a BigQuery and can go straight to the source. This is handy if you are looking for certain things over and over or you need to reference some piece of data often.
Its a nice structured way to pull data without a local RAG system or even NotebookLM. This is especially useful for data that might get updated over time, like census data or financial statements, you can ensure you always get the most current data.
Are there a million ways to do this type of thing already? Sure.
Is it the best way to do it? I don't know
Is it a viable way to do it and slots nicely into ICM? Heck yeah.
I will need to dig into this more. This might change the way I build my councils going forward for when I build an advisor off a real person. Instead of distilling them down into 2 or 3 md files, I can just point the LLM to interviews, papers, etc by said person to get a better responses.
4
3 comments
Rich C
5
Google OKF and ICM
Clief Notes
skool.com/cliefnotes
What we give away free beats most paid courses. Build durable AI systems with a Marine vet and Edinburgh researcher. 40+ lessons, growing.
Leaderboard (30-day)
Powered by