Running local LLM with outsourced web search knowledge

A Reddit user asks if it's possible to run a lightweight local model focused on reasoning while outsourcing its knowledge to web searches. The post has 30 upvotes and 36 comments discussing approaches.