The Missing Leg — A Framework the Vault Held Two-Thirds Of
Iccha-Kriya-Gyana Shakti Tri-Cluster for Content, built 2026-05-14 from a synthesized corpus, against the teacher's own account, ingested 2026-08-03.
Iccha-Kriya-Gyana Shakti Tri-Cluster for Content, built 2026-05-14 from a synthesized corpus, against the teacher's own account, ingested 2026-08-03.
The May page is accurate. Every rasa is in the right cluster. It names Śānta as the home base of Icchā and Karuṇā as the foundation of Kriyā.
And it stops. It never identifies the third base rasa, because the corpus it was built from didn't contain one.
Pashupati: adbhuta is what makes you noticed, hāsya is what makes it stick — and these are the two key secrets to gyāna-śakti, which all is on a base of śṛṅgāra. So you can see there's a base for everything.
This is the vault reading itself, and the finding is uncomfortable in a specific way: nothing looked wrong.
The page had no error. It had a gap shaped like nothing — an absence with no marker, because a two-thirds pattern doesn't announce that it's partial. Two bases read as two observations. Three bases read as a doctrine.
And the doctrine is the valuable part. With the third leg in place:
Each base rasa is a reception. Each pair of working rasas is an emission. Śānta lets you see before Vīra and Raudra act on what's seen. Karuṇā lets you care about someone specific before Bībhatsa and Bhayānaka move them. Śṛṅgāra opens the senses before Hāsya and Adbhuta express. The receiving must run first or the emitting has nothing to work from.
That sentence is not in the source either. It's only available once all three are visible — which means the completed framework yields something neither the teacher nor the synthesis states.
The failure the vault's own procedures didn't catch. The May ingest ran correctly. The page was well-built, well-cited, and cross-linked. Nothing in a lint pass, a hub audit, or a ghost-link sweep detects an argument that is missing its third instance, because every existing part is sound.
Structural incompleteness is a distinct defect class, and the vault has no test for it.
The existing QA apparatus checks: do links resolve, do sections exist, are claims cited, are contradictions preserved. All of those are local — they check parts against a standard.
Nothing checks whether an argument's shape is complete. A page asserting a two-part pattern where the source has three passes every test the vault runs.
The candidate test is cheap and specific: when a page names a pattern with N instances, ask whether N is the source's number or the page's number. Two-of-three is the dangerous case, because two instances already read as a pattern — which is precisely why nobody looks for a third.
Extension worth considering: the same defect should be more likely in synthesis-derived pages than in primary-source pages, because a synthesiser compresses and a compression that drops one of three items leaves no trace. Which would make the vault's `source: corpus synthesis` pages a natural audit population.
- The defect should recur. Audit a sample of `source: corpus synthesis` pages against their primary sources where those exist — do other two-of-three patterns turn up?
- The N-check has to be runnable. Test whether "is this the source's count or the page's?" can actually be answered without re-reading the whole source; if it can't, it isn't a usable lint rule.
- Counter-case: a synthesized page that correctly carries a three-part pattern. If synthesis usually preserves structure, this is a one-off rather than a class.