Skip to main content
Derived generation turns selected Raw evidence into a concise, reviewable Markdown candidate. Use it to organize accurate source material for retrieval and conversation—not to create missing experience. Every generated candidate starts Private and keeps source provenance. It becomes usable by runtime workflows only after you review, save, mark it Public, publish a version, and activate that version.

​
Generate a candidate

1

Select up to 20 Raw documents

Choose only the sources needed for one topic. Focused selections make factual review easier.
2

Choose a preset or custom instruction

Ask for a specific artifact, such as project context, a role summary, or an explanation of a decision and its constraints.
3

Choose the exact model and target path

TLDT uses the model you select. The output receives a target path and keeps links to its Raw sources.
4

Review the Private candidate

Check every date, actor, metric, denominator, time range, claim of ownership, outcome, limitation, and uncertainty.
5

Save through the normal document flow

Keep the candidate Private until you are comfortable making it eligible for publication.

​
Generation limits

Generation does not:
  • publish or activate a Memory version;
  • mark a document Public;
  • invent missing facts or strengthen weak evidence;
  • convert team work into individual ownership;
  • silently choose a different model;
  • silently truncate selected sources;
  • reconcile contradictions across your entire Memory;
  • learn automatically from Fast Check or Simulate.
If the selected context is too large, TLDT returns a context-overflow error rather than silently dropping source material.
Generation helps with organization. The owner supplies authority through evidence, review, and explicit publication.