DefendableOS · Overview
What DefendableOS is
Section titled “What DefendableOS is”DefendableOS is the engine that verifies agentic work. It runs the referee — a deterministic rulebook engine that applies declared rules and throws flags. It never grades by opinion. It is not a judge model.
DefendableOS is the engine. DefendableCloud is the hosted proof vault. The engine runs the rulebook; the vault runs the receipts.
The rulebook lives in Flight Sheets — versioned, declared eval templates carried by DefendableCloud. The engine consumes a Flight Sheet + an agent submission and emits a deterministic verdict: which rules passed, which raised flags, what tier (low/mid/high), and what severity (honey/jelly/propolis). A human approves; the Cloud mints a hash-chained receipt.
Why it exists
Section titled “Why it exists”“Offense goes dark. The business is offline. It can’t score.”
Most AI platforms build offense — bigger models, faster agents, prettier dashboards. Few build defense. We build defense. When AI offense fails — and it does · constantly — DefendableOS is the layer that says what happened · why · what failed · what got repaired · and what trains the next model.
The three things the engine does
Section titled “The three things the engine does”- Run the rulebook. Each Flight Sheet declares the rules: structure, schema, math re-derivation, evidence requirements, policy gates. The engine applies them deterministically.
- Throw flags. A check passes or raises a flag. Each flag has a tier (low/mid/high) and contributes to severity (honey/jelly/propolis). Score = % of declared rules satisfied.
- Sort the failure. Every flag falls into one of three buckets: work-defect (math/schema/evidence — the agent missed; fixable, resubmit), deal-finding (policy gate failed — the math is right, the rule says no; not a rework), stack-fit (model/compute below the lane — bigger brain, not a resubmit).
That’s the entire engine doctrine.
What you read next
Section titled “What you read next”- What It Is — plain English · principal-grade.
- What It Is NOT — disambiguation from generic AI SaaS, judge models, and Web3 token projects.
- Rulebook Engine — how the referee actually works (deterministic checks, structured executor, math re-derivation, DSL gates, variable penalty).
- Why Now — the market timing argument.
- Buyer Profile — the 5 personas + 8 verticals.
- Use Cases — what the platform actually does in production.
- Pricing Philosophy — trust infrastructure pricing.
- Board Flight Sheet — (engagement-model doc · the CRE-broker pre-market plan, distinct from the eval Flight Sheets carried by DefendableCloud).
🐝 The engine runs the rulebook · to the shed.