I have been thinking about developing a Chat that talks like a friend who remembers everything about you .
It is in web app state now https://t.co/BQFpda8gfA
Opus 4.7 with medium effort itself is consuming lots of tokens . Not seeing a significant quality in the solutions . It missed to test if the existing services are working fine after creating a Google security layer .
Thinking of evaluating Codex next month .
I organized my entire product code knowledge base into a LLM Wiki and i have a bunch of markdown files organized under assets, resources, products etc. This helped me to create a slimmer CLAUDE.md file where it was 3.5k tokens before change and now just 1k tokens after using LLM Wiki.