Ramblin' & Amblin' (2026-10-22): Okay

Okay. I've also found telling the AI to leverage DuckDB and SQLite locally to build a strongly typed data lake and while this is obviously no replacement for persistent memory what it does is it gives you a foundation for being able to to store domain data that's relevant to you not just cache but you can start to build up a historical knowledge base of the different resources that you encounter so that way you can start to push the AI to leverage that versus always hitting APIs or like for example you know if you're on GitHub GitHub has it's the GitHub client which allows for connecting a read- only token to be able to run commands through GH API and then you can say okay SQLite and DuckDB data lake based off the GitHub history for this repo and you know the the schema for uh for our GitHub is is is is well documented and uh you can start storing some of that old data maybe you know stuff that uh is already inert like uh old PRs um uh common history things like that you start building that out in your local database and uh instead of making calls and and wasting uh API uh limits rate limits on calls upstream to GitHub you build that data locally and uh what you end up with is a portable uh history that um that um um lets you uh own your data a little bit more and not have to uh be dependent on the uh network connection for all your work and uh you can then connect to that database either through your uh console you this is the other part is that you build a interface uh to be able to connect to that schema and generate in uh schema definition document document and erd and a d relationship document so that way you can reason with it and then you can connect to it through your database tool and now you have a browsable data store that you can um like I use the uh dweaver um you can connect to that uh SQLite database and and browse and look through your history and then ask better questions about hey you know like I see uh I'm missing an association here between the common history and common history and pr resolution um maybe there's a um um opportunity for for streamlining that relationship I don 't know I'm just making up stuff here um um and then you're talking about modeling and updating your local data store uh with the uh with the API and and and AI instead of just fetching fetching and fetch ing and fetching and fetching the same data over and over again you build out your local repository and the uh and you tell the ad to build the tools to um process its its queries through that and uh ensure that it stays up to date and uh and uh that you have uh tools to be able to access this and you just describe your environment your environment is going to be different I'm like I know that sounds super vague but that's really where you start um because these tools will know your system better and what is available to you and make recommendations based off of what it knows the most about your environment so I'm trying to be super uh specific aside from duck db sql ite database I mean if you want to do um uh postgres or mysq l or pongo I don't whatever um the point is just having some place that you can reason with this data that it's not always live calls over the uh over the network and burning up your rate limits and leaving leaving you with really nothing uh uh to take away at the end of the day at least you have some data all right thanks bye