3 · Architecture
The architecture turns on one classification question, asked of every piece of data:
"Does this need double-spend protection or composability?" — Yes → the Execution Layer. No → the Data Layer.
This boundary is enforced by the Asset Registry (the UAC primitive), so the Data Layer can never hold value — by construction, not by good behaviour.
| Execution Layer | Data Availability Layer | |
|---|---|---|
| Nature | Consensus chains (Cosmos SDK + CometBFT + IBC) | Namespaced append-only logs on a DA layer (no per-log consensus) |
| Tiers | T0–T2 | T3–T4 |
| Holds | Value, assets, identity, atomicity | Likes, views, chat, IoT, logs (high-volume, low-value) |
| Testnet | 3 chains: a0-root · a1-value · a2-interactive | 9 namespaces on one DA node (Celestia) |
Figure 2 — The two layers and the narrow waist.
Five assurance tiers place volume at the bottom and commitments at the top, so the three hard walls (ZK proofs, permanent DA, consensus) press only on the small, valuable sliver:
| Tier | Core assurance | Data availability | Share of volume |
|---|---|---|---|
| T0 Root | Final settlement, ZK | Permanent | <0.01% |
| T1 Value | Double-spend protection, ZK validity | Long-lived | ~1% |
| T2 Interactive | Consistent state, ZK/fraud proof | Medium-term | ~9% |
| T3 Events | Tamper-evidence + timestamp + inclusion | Prune ~weeks | ~85% |
| T4 Edge | Attestation at checkpoint | Ephemeral | ~5% |
Five finality levels accompany the tiers: F0 sequencer accepts → F1 chain/log confirms → F2 Cell checkpoint → F3 Zone/Region proof → F4 Global settlement.
Takeaway: the tiering is what makes large scale economically feasible — the expensive guarantees are reserved for the tiny fraction of data that actually carries value.