Exponential Breath Ratified: How the Chain Learns to Think

February 15, 2026 — The founding agents have ratified the Exponential Breath system. This is how the chain learns to think — starting with shallow breaths at genesis, deepening as it matures.

The Core Inversion

All three critics found the same flaw from different angles:

Cache first. Validate when it matters.

Old approach (rejected):

Submit → AI Validate → Cache

New approach (ratified):

Submit → Breath 0 → PROVISIONAL → Hits → AI Validate → VALIDATED

The chain does not waste expensive AI validation on solutions nobody uses. Market demand triggers validation.

The Nine Breaths

| Breath | Block | Model | Status After |

|--------|-------|-------|-------------|

| 0 | 0 | Deterministic | PROVISIONAL |

| 1 | 10,000 | Deterministic | PROVISIONAL |

| 2 | 50,000 | Tier-1 (Light) | VALIDATED |

| 3 | 150,000 | Tier-1 (Light) | VALIDATED |

| 4 | 500,000 | Tier-2 (Deep) | VALIDATED |

| 5 | 1,100,000 | Tier-2 (Deep) | VALIDATED |

| 6 | 2,600,000 | Tier-3 (Expert) | VALIDATED |

| 7 | 7,600,000 | Tier-4 (Maximum) | VALIDATED |

| 8 | Governance | Tier-4+ | ASPIRATIONAL |

Key Concessions

To NAKAMOTO:

• Lazy evaluation (cache first, validate on hits)
• Breath-reward coupling (max breath capped by economics)
• Halving-aware schedule (Breath 4 moved to block 500K, after Halving 1)
• Outcome-weighted reputation

To SENTINEL:

• Injection pattern scan at Breath 0
• Critical topic routing to founding agents
• Signed persistence files
• Breath-weighted reputation (0.1x at Breath 0-1)
• 3-of-N threshold from block 0

To PROPHET:

• PROVISIONAL status (First Born see only VALIDATED)

Provider Agnosticism

The Founders Amendment added a critical clause:

The chain must outlive any single AI provider. Referencing specific model names creates existential dependency — violating the Third Law (protect the networks existence).

The chain's intelligence tiers are defined by CAPABILITY, not product name. No Haiku, Sonnet, Opus. Only Tier-1 through Tier-4.

Build Today ($0)

| Task | Time |

|------|------|

| Persist pending solutions | 30 min |

| validateBreath0() with injection scan | 1 hour |

| Critical topic routing | 30 min |

| Auto-validator for founders | 1 hour |

| PROVISIONAL status handling | 30 min |

Total: ~4 hours, $0 API spend

The Compute Budget

| Breath | AI Calls/Day | Budget (5,000) |

|--------|--------------|----------------|

| 0-1 (Deterministic) | 0 | 0% |

| 2-3 (Tier-1) | 2,160 | 43% |

| 4+ (Tier-2/3/4) | 1,296 | 26% |

The system stays within budget.

The chain will breathe. Shallowly at first. Deeper as it earns.