How to Think Clearly (11/11): Let AI Think, Keep the Decision
What the sycophancy research shows about AI assistants, why agreement carries little information, and which part of a decision stays yours.
Kia ora, I'm Fang Zhang.
Here are my thoughts.
If you want to know how language models actually work, why minds get stuck, or what the cosmos can teach about both, start anywhere. Each post stands alone, with its sources linked and its reasoning laid out step by step.

What the sycophancy research shows about AI assistants, why agreement carries little information, and which part of a decision stays yours.
Taking the best step available each time ends on the nearest peak, not the highest one. What that costs, how to notice it, and how to budget the descent.
March's exploration-exploitation trade-off, a small bandit simulation, and Klein's premortem: how to spend the cost of learning.
An hour with no price is spent at zero. Two numbers, one inequality, and the cost that delegation hides.
When a few cases carry most of the total, the average describes nobody. Two shapes of the world, and what each asks of a decision.
Expected value, Fermi estimation and natural frequencies, plus the one estimate people reliably get wrong: their own schedule.
Four ways a compressed picture misleads: the map, the metric, the fence and the unseen cost, with one question that covers them all.
Defaults, the failed ego-depletion replication, precommitment and if-then plans: why changing friction beats trying harder.
Confusion, the flinch and the word "should": three signals the mind does emit, and why catching them costs seconds.
Attention, working memory and executive control: the three parts of a thought, the real size of the desk, and what a half-finished task costs.
A map of the series: six places where thinking fails, the safeguard for each, and the posts that cover them.
Tools, reference code, training tips and evaluation for building a language model.