Handbook
General vs specific intelligence
Operators choosing how a session starts, and authors deciding whether logic belongs in the general catalog or a domain pack.
Updated
Who is this for?
Operators choosing how a session starts, and authors deciding whether logic belongs in the general catalog or a domain pack.
What problem does it solve?
Not every decision is domain-specific. General intelligence answers which problem class, ladder, and budget apply before a domain pack runs. Specific intelligence answers what deterministic decisions and legal LLM accelerators exist for that problem family.
What can you do?
| Mode | You can… |
|---|---|
| General | Resolve problem class, reasoning ladder, token budget, and pack selection from the catalog. |
| Specific | Run per-domain packs with policies, flows, and fixtures tuned to one problem family. |
Typical flow: general intelligence selects policy → specific pack executes under that policy → session records both.
What can you not do?
- Run specific packs without a problem-class context — general intelligence establishes the session frame.
- Duplicate LCDL runtime inside a pack — packs compose existing governed tasks and flows.
Limitations
The general catalog and domain index evolve in waves. Some domains are documented but not yet at operator-ready depth. See Packs and domains.
Next steps
- Packs and domains
- Reasoning economy
- System depth: paired system page