Your analytics tells you what happened. Rungs tells you why — and what to do. Three questions no other tool can answer: Why did this happen? What will happen if we change X? What would have happened if we'd done Y? No math. No jargon. Just the business case.

Direct answer to "97% of the value is in the LLM." Side-by-side comparison of Claude alone vs Rungs alone vs both together on the same question. Visual benchmark proof. Why the engine is the IP and the LLM is the front door.

CYP2C9 pathway drug interaction traced. $15K database check prevents dangerous Warfarin-Fluconazole bleeding events.

Texas grid failure. 80% gas infrastructure, not wind. $300M fix prevents $200B loss. Causal attribution overturns political narrative.

345kV transmission cascade traced to a single Zone 2 relay misconfiguration 6 months prior — not the initial fault. NERC CIP compliance gap → N-1 contingency becomes N-2 blackout. Counterfactual: correct relay config prevents full cascade. $1.2B loss avoided.

Program evaluation. $800K breakfast program (8% budget) outperformed $4.2M software (42%). Causal isolation prevents $4.2M misallocation.

For leaders and strategists. The 30-year unsolved problem in business intelligence. Why every major decision is a causation question — and why every current AI tool answers it wrong. No math required.

How Rungs works, what the 99.15% score actually means, and what it costs to deploy at scale — explained without jargon. The 3 questions standard AI cannot answer. Real examples across 4 industries.

Mark's build log: Causal Red Team just shipped (5-attack adversarial stress-test, grade A/B/C/D), full list of what's in the engine, 2 features in progress, and 8 edge roadmap features that aren't in any open-source tool today — Transportability, Fairness Auditing, Network Interference, Federated Causal Inference, and more.

Complete technical review package for Brian and Scott. 80-assertion test suite with all DGPs embedded. Every claim is independently verifiable: grep the imports, run the tests, check the architecture. Includes Causal Red Team — a 5-attack adversarial stress-test that grades any ATE A/B/C/D. No trust required. Clone the repo and run python3 test_all.py.

Addresses whether Rungs depends on Claude, what minimum LLM the NLP layer needs, and whether building a custom LLM makes sense. Includes live training status: 7,736 examples, fine-tuning Qwen2.5-1.5B with LoRA right now on Mac M-series. After training: $0/query, no API, full Rungs IP end-to-end. Covers two modes, 32-point benchmark lead over GPT-4, build-vs-borrow analysis, and full claim-by-claim verdict.

Addresses the regex/NL parser question directly. Explains the CLadder 96.96% result and what it means for hallucination. Clarifies the architecture — what the parser is vs. what Rungs is — and where to focus.

10 shipped · 5 in progress · 15 planned. Two tracks: Rungs engine improvements and consumer recovery sites (8 domains). Updated as work ships.

AI-powered day/week/month organizer built on Rungs' causal reasoning engine. Understands task dependencies, cascades schedule changes, anchors to goals. $58B market, 200M target users. Thorough analysis: product vision, competitive landscape, revenue model, GTM strategy, market projections.

We proved it: Rungs-supervised fine-tuning takes a small LLM from 50% → 93.4% — within 1.4 points of human expert, beating GPT-4o by 16 points. The business: companies bring their model, we return turbocharged weights. Marketing roadmap, pricing, GTM strategy, and the Crazy Eddie video script.

The simple story: what we built, what the numbers mean, how the business works. Written for Scott and Brian — no jargon, no assumptions. Analogies: car tuning + teenager becoming an adult. Includes pricing, how a deal works, FAQ.

Marketing site concept in Fallout / Vault-Tec aesthetic. Green phosphor terminal, CRT scanlines, S.P.E.C.I.A.L. stats, before/after benchmark comparison, Wasteland Leaderboard, testimonials from Diamond City, and the full pricing table. "War never changes. But your AI can."