Managing context in JSON/Markdown files for near-zero LLM switching costs
August 7, 202552 words
Connor Grant
@cmgrant0
If my hunch is right, managing your own context (knowledge) in JSON/Markdown files brings the switching cost for LLMs to near-zero
I've been going hard on beefing Claude up for the past few months but I'll probably move to GPT5 if they drop a CLI agent