Why Manual Chain-of-Thought Hurts o1, o3 and o4-mini

OpenAI's reasoning best practices say that prompting a reasoning model to think step by step or to explain its reasoning is unnecessary, because these models reason internally. OpenAI adds that such instructions may not help and can sometimes hinder the model, so a manual Chain-of-Thought instruction ("Think step by step, break the problem into sub-tasks") is best left out of o1, o3 and o4-mini prompts. State the goal, the constraints and the output you want instead (a Goal-Oriented Specification).

Developer: Solve this SQL reconciliation problem.
- Problem Definition & Schema: {{SCHEMA}}
- Input Constraints: Must handle duplicate timestamps and null foreign keys without window-function memory spills.
- Desired Output Schema: Return only the validated SQL query inside a single code block.