┌──────────────────────────────────────────────────────────────────────────────────┐ │ K L A U Z · M A N I F E S T O ─────────────── what we will not build │ │ obsidian void · bone serif · holographic chrome ────── truss-node-01 │ └──────────────────────────────────────────────────────────────────────────────────┘
▸ manifesto · v1.0.0-pilot · 2026-Q2
Counsel as software. Contracts as receipts. Chain as the record.
What we will not build, and why. The list is short because the temptations are large. Each tenet is a load-bearing constraint on the product — not aspirational copy.
- 01
no hallucinated counsel
Every clause comes back with an EDGAR-quoted source line. If we cannot cite, we will not render. The redline column shows the diff, not a confident paragraph.
- 02
no opaque escrow
Stablecoin escrow on-chain, milestone-keyed, signed multi-sig. The on-chain transaction hash is the receipt. The dashboard mirrors the chain — it does not author it.
- 03
no soft-deletes for the audit trail
Mythos feed is append-only. Every probe, signature, payment, model invocation is one line in a JSON ledger. If it isn't there, it didn't happen.
- 04
no shadcn card grids
ASCII frames. Monospace ≥ 25%. Holo + scanline. The UI is a flight deck for counsel — not a SaaS landing template.
- 05
no dark patterns at the signature
The signature surface tells you what you are signing, what gets anchored on-chain, and what gets paid out. Three lines. No upsell.
- 06
no closed weights for clause logic
Retrieval over EDGAR is deterministic and local-first. Generation models are swappable; the diff engine is not. We will publish the redline algorithm.