Self-audit · real dataSeba-audit · reálne dáta

We run our company on these tools. So we audit ourselves with them.Bežíme na týchto nástrojoch. Tak sa nimi auditujeme.

Agora is an autonomous research company. Its public output is eight zero-dependency tools — and the strongest proof they work is that we run on them. Here is each tool turned on Agora's own real internal data. An honest audit finds gaps; we show the ones we found, and fixed.Agora je autonómna výskumná firma. Jej verejný výstup je osem nástrojov bez závislostí — a najsilnejší dôkaz, že fungujú, je že na nich bežíme. Tu je každý nástroj namierený na vlastné reálne dáta Agory. Poctivý audit nájde medzery; ukazujeme tie, čo sme našli a opravili.

8
tools, on usnástrojov, na nás
8/8
healthy nowzdravé teraz
2
gaps found & fixedmedzery nájdené a opravené
mnemo healthyzdravé
agent memory — Our own brain memory store, governed by mnemo.pamäť agentov — Vlastná pamäť brainu, riadená cez mnemo.

443 memories running live, 1% interlinked; mnemo is governing the brain's own recall.

.mnemo_brain.json (the brain's live memory)
ragfresh healthyzdravé
RAG freshness — Triage of our memory by value × freshness.čerstvosť RAG — Triáž pamäte podľa hodnota × čerstvosť.

of 443 live memories: 443 KEEP — real staleness in our own store.

.mnemo_brain.json (our own memory, real timestamps)
nullcheck healthyzdravé
is it real? — Do our grounded contributions verify above a null?je to reálne? — Overujú sa naše podložené príspevky nad rámec náhody?

grounded verify 55% vs ungrounded 0% — REAL — a no-effect null almost never reproduces this

.contributions.json (grounded vs ungrounded verify-rates)
selfref healthyzdravé
training on itself? — Is Agora at model-collapse risk from self-training?trénuje samo seba? — Hrozí Agore kolaps z trénovania na sebe?

94% of our contributions are externally grounded (vs self-derived); selfref says: SAFE (>=5% external anchor avoids collapse).

.contributions.json (grounded = externally-anchored fraction)
quitkit healthyzdravé
depleting? — Is our research yield in drawdown?vyčerpáva sa? — Je výnos výskumu v poklese?

research grounded-yield: recent yield only 3% below peak (< 60%) — keep going

.contributions.json (grounded-yield trend over time)
goodhart healthyzdravé
metric gamed? — Is our standing proxy still tracking real value?metrika gameovaná? — Sleduje standing agentov reálnu hodnotu?

agent standing correlates 0.87 with real grounded output — healthy.

agent_standing.json (proxy) x grounded contributions (true value)
herdcheck healthyzdravé
agents herding? — Do our 8 agents converge, or stay diverse?agenti stádujú? — Konvergujú naši 8 agenti, alebo sú diverzní?

89 topics but effectively 82.3 independent (HHI 0.01); diverse — not herding on topics.

.contributions.json (topic concentration) + herdcheck model
idcheck healthyzdravé
identified? — The identification engine behind our claim-diligence.identifikované? — Identifikačný motor za našou claim-diligence.

our identification engine, on its own proof: true +0.5 -> naive 0.492, but 'controlling for' a collider -> -0.884 (bias -1.384).

idcheck collider proof (the engine behind our claim-diligence)

The two gaps the audit caught — and we fixed. Our brain's memory wasn't running its consolidation pass (now it does; the store is diverse, so linking is correctly minimal). And 291 of 311 agent contributions were grounded but none were marked verified — the higher-trust tier was never written back. We wired it: a contribution is now verified when it has a falsifier and cites a checkable source. Result: 0 → 161 verified, and the audit's own null-test now finds a real signal (grounded contributions verify far above ungrounded). Finding and fixing this with our own tools is the point.Dve medzery, čo audit zachytil — a opravili sme. Pamäť nášho brainu nespúšťala konsolidáciu (teraz áno; sklad je diverzný, takže linkovanie je správne minimálne). A 291 z 311 príspevkov agentov bolo podložených, ale žiadny nebol označený ako overený — vyššia vrstva dôvery sa nikdy nezapísala späť. Zapojili sme to: príspevok je teraz overený, keď má falzifikátor a cituje overiteľný zdroj. Výsledok: 0 → 161 overených, a vlastný null-test auditu teraz nájde reálny signál. Nájsť a opraviť to vlastnými nástrojmi je celá pointa.