The internals,
on display.
Most AI demos are a chat box and a chart. These aren't. Each scene visualises the actual cryptographic, consensus, or verifier internals of a running system — deterministic data, real signatures, real Pedersen commitments. Pre-computed from the Banterpacks pipeline, rendered here. No mocks.
- 01
The reasoning step, witnessed five times
A single CoT step flows through T1 verifier → T2 judge → T2.5 self-correct → T3 panel → Enforcer. Real verdicts. Real rewinds.
TDD-012 §3, §8, §17 - 02
Four ways to think about the same prompt
Analytical, Creative, Adversarial, Domain Expert — four cognitive styles inspecting one task with their own structural logic. ELO-weighted, meta-controller-tuned.
TDD-005 cognitive layer - 03soon
Proving alignment without revealing the score
Pedersen commitment over the similarity score. Schnorr OR proofs over the bits. The verifier learns the step is above threshold and nothing else.
TDD-005 zk.rs · P102.9 · P102.18.2 - 04soon
Four replicas, one decision
Pre-prepare, prepare, commit. Ed25519 signatures collecting into quorum. Equivocation detection if a replica lies.
TDD-005 bft.rs · P102.18 audit-fix - 05soon
Every action, signed and chained
Tool action → SHA-256 → Ed25519 signature → Merkle node → root. Cold-tier archive. Audit trail you can verify in another language.
TDD-004 provenance