Compose · Analyze
Compose verb → Lodestar / grounding kernel
The Compose verb realized as Lodestar — it finds the minimal (~1%) record set whose meaning explains an entire corpus, the grounding kernel, usable as both an index and an answer path.
Out of a million records, which subset would you keep if you could only keep the few that, together, still explain the whole collection?
The grounding kernel — roughly 1% of the records. Lodestar (the Compose verb) finds that minimal load-bearing set, the few stars that, if you knew only them, you'd still understand the sky. It doubles as an index (search the 1%) and an answer path (ground answers in it).
Source: calyxdocs/12 (Lodestar)
Out of a million records, which subset would you keep if you could only keep the few that, together, still explain the whole collection?
The grounding kernel — roughly 1% of the records. Lodestar (the Compose verb) finds that minimal load-bearing set, the few stars that, if you knew only them, you'd still understand the sky. It doubles as an index (search the 1%) and an answer path (ground answers in it).
The few stars that explain the sky
The Compose verb is realized as Lodestar. It asks the most ambitious question in the system: of all the records in a corpus, which minimal set — often around 1% — carries enough meaning to explain the rest? That load-bearing subset is the grounding kernel.
A kernel is both an index and an answer
The kernel is useful twice over:
- as an index — searching the 1% is far cheaper than searching everything, and the kernel was chosen precisely because it covers the space;
- as an answer path — because the kernel is what explains the corpus, grounding an answer in it is grounding it in the corpus’s own backbone.
Lodestar builds on the verbs beneath it: it needs Differentiate to know which records carry real bits, and it serves grounding by giving answers a compact, trustworthy thing to stand on. The kernel is also the seed of the bigger thesis — a grounding kernel at any scope, plus the guard and the oracle, is what the Calculus of Association frames as a grounded intelligence substrate.
Practise to master it
Sign in to practise — practice items are graded per learner and update your mastery toward the lesson gate.
Where this sits in the concept graph
Leads next
- Build an AI-memory app on Calyx
- Oracle / consequence prediction
- Toward AGI — the kernel as substrate
- Ward / fail-closed guard
Related
- Aster (LSM storage)
- Count verb → Loom / associations
- Differentiate verb → Assay / signal bits
- Store meaning, not tokens
- Symmetry of Knowing
- The four verbs
- Toward AGI — the kernel as substrate
Sources
- calyxdocs/12 (Lodestar)
- calyxdocs/17
- PRD 08_LODESTAR