AI Lab · by Yui Tanaka · 2026-10-20

RAG vs long context: which do you actually use?

Long context windows keep growing, but retrieval still wins on cost and freshness in my projects. What does your stack look like?

41 sparks · 20 comments

Comments

Isha Menon · 12 likes

Open weights or it didnt happen

Haha my thoughts exactly

My GPU just cried reading this

Arif Hasan · 0 likes

Good take, hadnt considered that

Aoi Ito · 8 likes

Benchmarks look great but real world usage is where it falls apart for me

Good take, hadnt considered that

This ^

Context window claims are always optimistic, test it yourself

This ^

The cost per token matters more than the leaderboard imo

Depends but yeah mostly

This ^

Amir Tan · 2 likes

Running a quantized version locally and its surprisingly usable

I read that twice and still chuckled

Fair point

Evals are getting saturated, we need harder ones

Good take, hadnt considered that

Narin Chai · 1 likes

This ^

Prompt sensitivity is still the biggest pain in production

Haha my thoughts exactly