ResearchOS journal

Research is a process, not one prompt

Published August 28, 2026

A single prompt can be useful for brainstorming, but it is a weak model for research. A meaningful question often contains several smaller questions: what happened, why it happened, what evidence supports competing explanations, and what important information is still missing.

The first step is planning. Breaking a broad question into focused objectives makes the research task more deliberate. It also gives the researcher a way to judge coverage later. If an objective has no reliable evidence, that gap should be visible rather than hidden by a smooth summary.

Next comes gathering. Tools can search connected sources and retrieve relevant material, but tool output must be treated as input to evaluate, not as a fact to repeat. Each useful result should be checked for relevance, provenance, and enough context to support a claim.

Evaluation is where research becomes more than collection. Evidence can support a claim strongly, weakly, or not at all. Sources can conflict. A responsible report records limitations and avoids suggesting certainty that the evidence does not justify.

Synthesis comes after that work. It is the act of connecting supported evidence into an explanation, comparison, or argument. Good synthesis stays close enough to the sources that the reader can review it, while still helping the reader understand what the evidence means together.

Writing is the final stage of this process. When a paper is drafted from planned objectives, selected evidence, and an inspectable trace, the result is easier to defend, revise, and learn from than a generic response produced in one step.