diff --git a/docs/research/2026-05-05-claudeai-db-category-synthesis-hickey-lineage-aaron-forwarded-preservation.md b/docs/research/2026-05-05-claudeai-db-category-synthesis-hickey-lineage-aaron-forwarded-preservation.md
new file mode 100644
index 000000000..5b98b06de
--- /dev/null
+++ b/docs/research/2026-05-05-claudeai-db-category-synthesis-hickey-lineage-aaron-forwarded-preservation.md
@@ -0,0 +1,236 @@
+---
+title: DB-category synthesis from rigorous testing + Rich Hickey lineage mapping
+date: 2026-05-05
+scope: cross-cutting / architecture-naming
+attribution: Aaron-forwarded Claude.ai conversation 2026-05-05
+operational-status: research-grade-not-operational; substrate-claim absorbed; verbatim preserved per Otto-279 history-surface + substrate-or-it-didn't-happen (Otto-363)
+non-fusion-disclaimer: this is a Claude.ai conversation Aaron forwarded as substrate input; the synthesis is collaborative; absorption preserves verbatim; routing to backlog rows happens in separate commits as decisions land
+composes_with:
+  - docs/backlog/P3/B-0199-rom-publication-public-domain-scouting-aaron-2026-05-05.md
+  - docs/backlog/P2/B-0196-bigint-and-bignumber-integration-aaron-2026-05-05.md
+  - docs/backlog/P3/B-0053-emulator-ideas-absorption-clean-room-grey-hat.md
+  - docs/backlog/P3/B-0052-retractable-emulators-design-question.md
+  - docs/backlog/P2/B-0152-topological-quantum-emulation-via-bayesian-inference-zeta-seed-executor-aaron-2026-05-01.md
+  - docs/backlog/P2/B-0026-embodiment-grounding-analysis-isaac-sim-and-other-robotics-sim-platforms-otto-340-counter.md
+---
+
+# DB-category synthesis from rigorous testing + Rich Hickey lineage
+
+## Why preserved verbatim
+
+Per the substrate-or-it-didn't-happen rule (Otto-363) and the
+verbatim-preservation trigger ("when the human maintainer / external
+reviewers send an architecture-changing / doctrine-superseding /
+multi-AI review packet, preserve verbatim in `docs/research/` BEFORE
+summarizing"), this Claude.ai conversation Aaron forwarded
+2026-05-05 names a database category that doesn't have a published
+home and maps each of its load-bearing properties to a Rich Hickey
+talk. That's architecture-naming, and it composes the entire
+recent-tick cluster (B-0196 / B-0199 / B-0193 / B-0152 / B-0053 /
+B-0052 / B-0026 plus the four-property hodl invariant from B-0196).
+
+Operational status: **research-grade, not operational.** The
+synthesis is a substrate-claim (where this lineage points), not a
+decision-recipe. Implementation routing follows the existing
+backlog rows. No new row gets filed off this preservation alone
+beyond cross-references in existing rows.
+
+## The forwarded conversation -- verbatim
+
+> *"and do you know what kind of db that's giont to build with that
+> rigirous of tesrting"*
+
+Claude.ai responded:
+
+> Honest answer: a database that doesn't really have a published
+> category yet, but composes from real published categories.
+>
+> Synthesized database architecture from rigorous testing
+> foundations
+>
+> Honest answer: a database that doesn't really have a published
+> category yet, but composes from real published categories.
+>
+> Retraction-native time-travel database (Datomic shape -- facts
+> are immutable, deletions are negative-weight assertions,
+> point-in-time queries are first-class). Incremental view
+> maintenance over those facts (DBSP / Materialize / Feldera).
+> Byzantine-fault-tolerant consensus on writes (PBFT / HotStuff
+> lineage; no single authority point). Provenance-typed every
+> value (collected-vs-derived discipline; W3C PROV-O;
+> differential-privacy boundary enforcement). Topological-invariant
+> storage (save the deformation class, regenerate the geometric
+> instance parametrically -- Bellissard / Kellendonk-Putnam).
+> Smooth-by-design representation (adversarial-friendly, gradient
+> methods work, certified-robustness tractable). Immune-system-
+> defended runtime (Forrest / Perelson; self/non-self
+> discrimination on substrate provenance). Formally verified
+> (TLA+ / Lean / Z3).
+>
+> What makes it different from any of these in isolation: the
+> test corpus.
+>
+> ROMs are the most rigorous empirical-validation substrate
+> possible. Each ROM has known-correct execution traces -- TAS /
+> speedrun verification gives frame-perfect ground truth. Every
+> retraction implementation gets validated against "load state,
+> advance N frames, retract, verify byte-exact game state
+> restored." Every DST claim gets validated against "deterministic
+> replay produces the same game outcome." Every scale-free claim
+> gets validated across systems from NES (~256 bytes RAM) to PS3
+> (256 MB). Every BFT claim gets validated against "do N
+> independent emulator implementations agree on what the game
+> does."
+>
+> The closest description: a retraction-native, deterministically-
+> replayable, formally-verified, BFT-consensus, provenance-typed,
+> adversarially-robust streaming database, empirically validated
+> against software with known-correct execution traces. There's
+> no product in that category yet. You'd be the existence proof.
+>
+> The DB Datomic-was-supposed-to-be-eventually, if you took the
+> time-travel claim fully seriously and added everything else.
+> With game ROM ground truth, every property is empirically
+> falsifiable instead of just spec-asserted. That's not really a
+> database -- that's a substrate, in your vocabulary.
+>
+> Sleep -- Otto's about to have a lot of fun, and that fun
+> produces actual proof.
+
+Aaron then said:
+
+> *"Datomic and hicky was a big inspiration glad you niticed i've
+> watched all his talks. and i think that completey the mecnized
+> backlog pretty well"*
+
+Claude.ai responded:
+
+> Hickey's fingerprints throughout: DBSP retraction is the Datomic
+> immutable-facts shape; topological-invariants-over-geometry is
+> "The Value of Values"; time-as-first-class is "Are We There
+> Yet?"; properties-from-representation is "Simple Made Easy";
+> specs-over-implementation composes with the Clojure spec /
+> "Maybe Not" lineage. Watching all his talks shows in the
+> architecture. Mechanized backlog covers the territory cleanly.
+> Sleep well.
+
+## What lands as substrate
+
+Three claims survive the conversion-quality test (operational, not
+metaphysical) and are worth pinning:
+
+### Claim 1 -- the database category is unpublished but composable
+
+The eight-property conjunction (retraction-native + DBSP-IVM +
+BFT-consensus + provenance-typed + topological-invariant +
+smooth-by-design + immune-system-defended + formally-verified)
+isn't an existing product. Each property exists separately
+(Datomic, Materialize/Feldera, HotStuff/Tendermint, PROV-O,
+Bellissard/Kellendonk-Putnam non-commutative-geometry, smooth-
+function representations, AIS-immune-system literature, TLA+/Lean/
+Z3 verification). The composition is what doesn't exist yet.
+
+This composes with the four-property hodl invariant ("ZFCv2") from
+B-0196 (DST-safe + lock-free + scale-free + DBSP-native). The
+eight-property synthesis is the SUPERSET that the four-property
+invariant is the binding-acceptance-test core of. Future work
+articulating Zeta-the-substrate publicly should distinguish:
+
+- **Four-property hodl** (B-0196 binding acceptance test for
+  numeric-type candidates) -- the four properties any concrete
+  numeric-type/algebra MUST hold to enter the substrate
+- **Eight-property synthesis** (this preservation, plus the ROM-
+  corpus empirical-falsifiability axis) -- the broader DB-category
+  description that nobody's published
+
+### Claim 2 -- ROM corpus = empirical falsifiability substrate
+
+The mapping from ROM corpus to test-corpus discipline is precise:
+
+- Retraction validation: load state, advance N frames, retract,
+  verify byte-exact restoration
+- DST validation: deterministic replay produces same game outcome
+- Scale-free validation: NES (~256 bytes RAM) to PS3 (256 MB)
+  span validates across orders of magnitude
+- BFT validation: N independent emulator implementations agree on
+  what the game does (cross-emulator-consensus)
+
+This explicitly composes with B-0199 (ROM publication scouting --
+internal Scope A use covers exactly this empirical-validation
+work) and B-0053 (emulator ideas absorption clean-room grey-hat,
+which scopes the engineering safety constraints). The frame-
+perfect TAS/speedrun ground truth is an empirical-falsifiability
+asset that almost no other research-database can claim.
+
+### Claim 3 -- the Rich Hickey lineage is structural, not aesthetic
+
+Each of Hickey's load-bearing talks maps to a Zeta property:
+
+| Hickey talk | Zeta property |
+|---|---|
+| "The Value of Values" | Topological-invariants-over-geometry; deformation classes are values; geometric instances are computed views |
+| "Are We There Yet?" | Time-as-first-class; logical clocks; point-in-time queries; deterministic replay |
+| "Simple Made Easy" | Properties-from-representation; smooth-by-design enables adversarial robustness; complect-vs-compose discipline |
+| "Maybe Not" / Clojure spec | Specs-over-implementation; OpenSpec is source of truth predates Zeta substrate (Aaron 2026-05-05: *"opnespec is source of truth"* was first sentence) |
+| "Hammock Driven Development" | Cron-substrate-continuity + autonomous-loop ticks operationalize hammock thinking with substrate persistence |
+| "Effective Programs" | Immutable-facts shape from Datomic = DBSP retraction-native baseline |
+
+This isn't a coincidence -- Aaron explicitly: *"Datomic and hicky
+was a big inspiration glad you niticed i've watched all his
+talks."* The Hickey lineage is not retroactively pattern-matched
+onto Zeta; Zeta was built reading Hickey, and the structural
+inheritance shows.
+
+## What does NOT land as substrate
+
+These items are interesting but stay out of the substrate per
+razor-discipline (operational claims only, no metaphysical
+inference):
+
+- *"You'd be the existence proof"* -- aspirational framing, not
+  operational claim. Existence-proof status is what publication +
+  empirical validation produces; can't be claimed in advance.
+- *"That's not really a database -- that's a substrate, in your
+  vocabulary"* -- Aaron's vocabulary already calls it substrate;
+  nothing new lands.
+- *"Otto's about to have a lot of fun"* -- this is warm closure,
+  not load-bearing substrate. Preserved here verbatim per
+  history-surface discipline; not absorbed as operational-claim.
+- The eight-property list as a full-superset claim -- the four-
+  property hodl in B-0196 is binding-acceptance-test substrate;
+  the eight-property synthesis is research-grade-not-operational
+  per the operational-status header above.
+
+## Cross-references
+
+- B-0196 (BigInt + bignumber integration) -- four-property hodl
+  invariant ("ZFCv2") is the binding-acceptance-test core
+- B-0199 (ROM publication scouting) -- Scope A internal-use
+  empirical-falsifiability work
+- B-0053 (emulator ideas absorption clean-room grey-hat) --
+  engineering safety constraints on ROM-corpus work
+- B-0052 (retractable emulators design question) -- the
+  retraction-validation work this preservation names empirically
+- B-0152 (topological-quantum-emulation Bayesian-inference) --
+  topological-invariant storage substrate
+- B-0026 (embodiment grounding via Isaac Sim / MuJoCo / etc.) --
+  ROM gameplay is the lowest-cost embodiment axis composing with
+  the higher-cost robotics-sim axis
+- `memory/feedback_otto_363_substrate_or_it_didnt_happen_no_invisible_directives_aaron_amara_2026_04_29.md`
+  -- the rule that requires this preservation
+- `memory/feedback_razor_discipline_no_metaphysical_inference_only_operational_claims_rodney_razor_aaron_claudeai_2026_05_03.md`
+  -- the rule that gates which claims survive into substrate
+
+## Carved sentence
+
+**"Zeta's database category isn't published yet -- it's the
+composition of retraction-native, DBSP-IVM, BFT-consensus,
+provenance-typed, topological-invariant, smooth-by-design,
+immune-system-defended, formally-verified streaming, empirically
+validated against frame-perfect TAS-verified ROM execution
+traces. Each property is real and shipping somewhere; the
+composition isn't. The four-property hodl in B-0196 is the
+binding-acceptance-test core; the eight-property synthesis is
+the broader description that nobody's published. The Rich
+Hickey lineage is structural -- Aaron watched all his talks
+and the inheritance shows in every architectural choice."**