Discovered a far better way for retrival using Reasoning based DB that creates a Table of contents(TOC) divinding the document into nodes without loosing semantics of original document ,One framework was @PageIndexAI#buildinpublic
Spent last wee studying making RAG bots and vector databases mainly using @neondatabase (pgvector)
Created one RAG trained on basketball data with some of my profile knowledge.
Try here: https://t.co/ptNfJh1QWk
Something backend devs eventually realize:
Schemas exist in multiple places.
Validation schema
Model schema
Database schema
Understanding how they interact makes system design much easier.
One underrated skill in engineering:
explaining what you built clearly.
Spent some time refining project descriptions today and realized writing impact > listing features.
Did a small QA pass on a project today and tried looking at it in three ways:
• bugs
• UX friction
• “good to have” improvements
That small mental model actually surfaces way more issues than just looking for bugs.
Interesting realization today while reviewing code:
Testing strategy isn’t just about coverage.
Unit tests protect logic.
Integration tests protect system behavior.
Smoke tests protect deployments.
Different problems, different tools.
Backend question I was thinking about today:
Do you handle validation at the API layer (Zod) or rely mostly on database schema validation?
Personally starting to see the value of doing both — different layers, different guarantees.
Worked on an agri-tech system during Smart India Hackathon 2025 called Kshetrin.
Focused on combining AI recommendations with hardware-driven irrigation systems.
Reached the finalist stage — huge learning experience.