AI-fïed is a research journal in public.
It documents what actually happens when AI systems are built, tested, and pushed beyond demo environments.
Not prompt tricks.
Not recycled tool lists.
Not surface-level productivity hacks.
The focus is structural:
- Why AI feels intelligent but isn’t
- Where retrieval systems fail silently
- What hallucinations really are
- How tokenization shapes output
- When automation reduces clutter — and when it creates it
AI-fïed exists to examine the gap between how AI looks and how it behaves.
The goal is simple:
Make AI simpler — by understanding it properly.
Not by dumbing it down.
By exposing its mechanics.
About Abhinav
I’m Abhinav.
I started AI-fïed while learning AI systems from the ground up — not from hype, but from building.
This platform documents experiments:
- Building a local RAG system
- Observing retrieval failures
- Testing chunk strategies
- Studying how language models actually process input
I am not presenting finished answers.
I’m documenting how systems behave in reality.
Because AI doesn’t break loudly.
It works — and still misses the point.
Understanding that difference changes how you use it.
