TTP v4 The Moat
12 Defensive Rings · 5 App Surfaces · Mythos-Grade · Lucent Lens
Why this matters · April 2026
On April 7, Anthropic announced Claude Mythos Preview — a frontier model that has already found thousands of zero-day vulnerabilities across every major OS and browser, 99%+ still unpatched. Agentic attack cycle times have collapsed to ~22 seconds. Project Glasswing ($100M, 12 launch partners + 40 invited orgs) keeps that capability defensive for the allow-listed few. Most startups are outside that fence. TTP v4.1 is how ML Systems runs a Mythos-grade defensive posture without needing to be inside Glasswing — credential-gated, rate-limited, circuit-breaker-backed.
Core
ML SYSTEMS
The Keep — Lucent Lens
- Lucent Lens: Glow within to help humans
- 10 ecosystem surfaces protected
- Closed-loop construction fintech
- NAICS 236115 — Rhode Island
- Transparency Trust as fiduciary duty
- PI: the aura is the protocol
App × Band Pricing Matrix
| TTP Band | Design StudioCDA | Decon LabREAPER | Builder's Open HouseMIA | The PitPIT LORD | Builders CollectiveVERA |
|---|---|---|---|---|---|
| L0 Public RecordScore 0–10 · 1¢/query | FreeBrowse templates · speed decay active | Free ScoutView reports · no AI analysis | BrowseView listings · no cart | View OnlySee open pits · cannot bid | BrowseView directory · no verification |
| L1 ML VerifiedScore 11–30 · 5¢/query | $400Single design package | $250/assemblyAI analysis · recovery score | List + Buy10% seller commission | $5–$50Standard to Proven bids | Browse+View profiles · request reviews |
| L2 Full APIScore 31–60 · 10¢/query | $10KValue chain · full design package | $2K/structureFull plan + DEM export included | Sell: 5%Reduced comm. · priority listing | $100–$250Forged to Tempered bids | $5KVerification · peer review access |
| L3 Double-VerifiedScore 61–80 · 25¢/query | $10KPriority queue · Opus AI access | $3.5KFull suite + DEM exports | Sell: 0%Zero commission · featured seller | $500–$1KLucent to Sovereign bids | $10K10 architect reviews included |
| L4 Ontology LicensedScore 81–100 · 50¢/query | $25KCustodian tier · direct with Sal | $25KCustodian · patent + Sal-direct | $10K/moLogistics Partner · The 27 | $1 + 0.1%Custodian bid + succession fee | $25KCustodian verification |
Decon Lab at 10x scale — $250/assembly · $2K/structure · $3.5K.
Band × Model Router
| TTP Band | Score | Default Model | High Complexity | Cost/Query | RPM | Daily Ceiling | Burst |
|---|---|---|---|---|---|---|---|
| Public Record | 0–10 | Haiku | Haiku | 1¢ | 30 | $30 | None |
| ML Verified | 11–30 | Haiku | Sonnet | 5¢ | 120 | $120 | None |
| Full API | 31–60 | Sonnet | Sonnet | 10¢ | 600 | $600 | 20 / 5s |
| Double-Verified | 61–80 | Sonnet | Opus | 25¢ | 1,200 | $3,000 | 50 / 5s |
| Ontology Licensed | 81–100 | Opus | Mythos | 50¢ | 3,000 | $12,000 | 100 / 10s |
v4 → v4.1 Additions — Mythos-Grade Defense
Prompt Defense · Multimodal
Multi-layer AI input security across text, image, audio, and video metadata. Adaptive-attacker posture covers the Airia "Lethal Trifecta" — multimodal injection, agentic tool abuse, and MCP protocol-layer exploits.
- 25+ injection patterns (text + img + audio + video metadata)
- MCP protocol-layer checks for agent tool abuse
- 6 heuristic signals + semantic scoring
- Adaptive-attacker rapid-response cycle
- Constitutional-alignment embedded at prompt boundary
Autonomous Threat Intel
Aligned with Anthropic's $100M Project Glasswing framework. 22-second breach-window SLA. Post-Glasswing posture for startups outside the 52-org allow-list — Mythos-class threat modeling without requiring Mythos access.
- $100M Glasswing framework alignment
- 22-second breach-window SLA matched
- Mythos ceiling (SWE-bench 93.9% / USAMO 97.6%)
- 10 anomaly types cross-surface + CVE sync
- RSP v3 transparency · incident disclosure
Agentic Perimeter
Defense against autonomous attack agents operating at 22-second cycle times. Multi-agent defensive response, vendor-ecosystem threat modeling, and contractor account isolation — the direct architectural response to the April 22 Mythos third-party incident.
- Multi-agent defense (counter to IBM Autonomous Security pattern)
- Vendor-ecosystem threat modeling as first-class
- Contractor isolation · no shared API keys
- Per-vendor posture score · auto-revoke on anomaly
- Outermost shell — last line before The Keep