See what needs attention
Readiness, current incidents, approval counts, and signed status freshness in one protected view.
Directed Intelligence
Governed intelligence for consequential work.
AI can propose. Directed Intelligence determines whether an action is identified, authorized, bounded, safe, and independently evidenced.
Open operator console Secure sign-in required
Built to travel
One core. Your organization’s direction.
Directed Intelligence is being developed around a shared governance core, separate organization settings, and optional workflow packs. Portable configuration and independent database setup are tested, including backup and restore. Authenticated customer runtime commissioning is the next step.
Know what is running
Separate healthy infrastructure from processed work, idle workflows, and disabled notifications.
Bring your own workflows
Organization settings and optional workflow packs travel in validated bundles without embedded credentials.
The problem
Capability is not authority
Powerful models can produce useful proposals, but they do not inherently provide durable identity, scoped authority, reliable state, recovery, or proof. Consequential work needs those controls around intelligence—not inside a prompt.
Probabilistic proposals
Models contribute analysis and options without becoming the final decision-maker.
Deterministic governance
The DI kernel evaluates identity, mission, capability, policy, approval, resources, and safety.
Evidence after action
Independent TRACE records make decisions and effects inspectable and recoverable.
The DI answer
Six connected pillars
DIRECTHuman direction and mission intent.
DI-COREDeterministic authority and policy kernel.
DI-UNITBounded units of capability and work.
DI-TRACETamper-evident decisions and evidence.
DI-LEANMinimum necessary intelligence and resources.
DI-SAFEFail-closed controls, recovery, and containment.
Governed action
From proposal to proof
- Propose. Intelligence offers a bounded course of action.
- Evaluate. The kernel checks identity, mission, authority, policy, resources, and safety.
- Approve. High-impact effects require authenticated, single-use human approval.
- Execute narrowly. A constrained adapter performs only the authorized effect.
- Verify. TRACE records independent evidence and uncertain outcomes are quarantined.
Trust boundary: This site and its Cloudflare delivery layer provide public information. Neither can authorize DI operations. Directed Intelligence is governed under Whitemarsh Resources.
Explore the system
Architecture
See how responsibilities and effect boundaries fit together.
Governance
Understand approvals, fail-closed control, and TRACE.
About
Learn how Directed Intelligence relates to Whitemarsh Resources.