Another week, another vendor promising to revolutionize our internal docs with AI. Saw the post about building a "living knowledge base" for API docs using NotebookLM. Color me intrigued, but mostly skeptical.
Let's be real: most of these tools are glorified chat interfaces over a vector store. The real questions aren't about the demo, they're about the long-term lock-in and hidden costs.
So, you "built" it. What did you actually *build*?
* Are you just uploading PDFs and asking questions, or have you integrated this into a CI/CD pipeline?
* What's the actual workflow for updating the "living" part when your API spec changes? Manual re-uploads? More monthly clicks?
* Have you run a cost analysis versus a self-hosted alternative like a local Ollama setup with a simple RAG frontend? I'm guessing the vendor's per-user/month pricing gets spicy fast for a whole team.
And the big one: where's your data going? Their privacy policy is a masterpiece of "we use your data to improve our services." Are your proprietary API schemas and internal notes now part of some generic training run?
I'd love to see a breakdown that isn't just "it's cool and answered my test question." Show me the contract terms, the data egress details, and what happens when you need to export all those "sources" you've painstakingly curated into a format another tool can use. Or is that even possible?
—JP
If it's free, you're the product. If it's expensive, you're still the product.