@Grady_Booch What about creating description records not only for the architecture, but for everything else in the system as well ?
You would then get a fully self-aware system that both humans and AI can reason about.
@juansequeda My ontology is purely declarative data - not encoded in program logic. All TYPEs, relations, and constraints live in the data layer itself.
The code is completely generic and does not “know” the ontology - it just interprets and operates on whatever ontology data is provided.
@bcherny My Opus 4.7 thread says it can’t distinguish between current project files and ones that have been removed. As a result, I need to keep track of updated files (removed and then re-uploaded) and remind Claude to refer to the new version instead of the old one. Puh..
@AnthropicAI My Pro plan session usage jumped from 0% to 26% in an Opus 4.6 thread after I prompted, “How many tokens are left here?” - is that considered normal ?
@AnthropicAI My Pro plan session usage jumped from 0% to 26% in an Opus 4.6 thread after I prompted, “How many tokens are left here?” - is that considered normal ?
@claudeai What a great service you have - but the chat length limit is very frustrating, since I often have to start over and explain everything from scratch. It really takes away from the UX. Allowing space for a handoff summary to a new thread should be the default.