Are we over-engineering AI memory? (Markdown vs. Vector DBs for small datasets)

Hey makers!

Lately, I’ve been looking closely at how independent builders and small teams are managing AI knowledge bases. It feels like the default "industry standard" is to immediately reach for a complex RAG pipeline and a heavy, paid Vector Database.

But I'm starting to wonder if we are over-engineering this for 90% of standard use cases.

Vector DBs are incredibly powerful for massive scale, but for smaller or non-massive datasets, they can be expensive, complex to query, and act as complete black boxes. If a search returns a weird chunk, diagnosing it is often a nightmare.

Lately, I've been experimenting with a dead-simple alternative for smaller projects, and I'd love to know if anyone else is doing this:

1. Dump raw data into a local folder.

2. Use a strict markdown schema (like a CLAUDE.md or similar instruction file) to tell the LLM how to summarize and map the data.

3. Let the AI build a flat, interlinked wiki overnight with directional backlinks (e.g., Supports [[Topic A]] or Contradicts [[Topic B]]).

I've found this to be 100% human-readable, highly portable across AI tools, virtually free to host, and shockingly accurate because agents can actually traverse the file logic rather than relying on pure mathematical similarity scores.

I want to hear from the builders and technical founders here:

• Are Vector databases actually a mechanical necessity for smaller, non-massive datasets?

• Or are we defaulting to them because it's the trendy stack right now?

• If you aren't using a Vector DB for your AI features, how are you handling knowledge retrieval?

Let's talk shop in the comments!