# Research Paper understand
-- Can you summarize the paper in your own words in three sentences
-- Help me grasp the big picture here
-- How well does the method work, really?
-- What's bottom line? Cite relevant sections from paper

# Update mental model (better understanding)
1) belief update: "I used to think X mattered. Now I think Y is the real lever"
2) failure points: "Under what conditions does this idea fail?"
3) micro-exp: "For next 3 days, if I do instead of Y and observe Z."

nascent

# Macros
-- Under what specific conditions does test-time compute outperform pre-training scaling?
-- How does the difficulty of the prompt dictate the optimal allocation of compute?
-- What are the distinct <scaling> properties of (1) method 1 and (2) method 2
-- Why does Beam Search degrade performance on easier problems at high compute budgets?
-- How is 'question difficulty' estimated without access to ground truth during inference?

# Generalised questions as researcher (ex paper SAPO)
-- What "deficiency" in existing group-based methods motivates this work?
-- How does model architecture exacerbate the optimization challenge?
-- What is the "rationale" for <improvement / concept>
-- What "empirical" evidence supports the method is superior

-> Instead of asking "what do you think about xyz?"
-> replace "you" to a group / demography of people who's perspectic you value
-> "What would be a good group of people to explore xyz? What would they say?"