THE GOVERNED AI CONTROL PLANE

Intelligence moves.
Authority stays.

A decision architecture that separates what AI proposes from what systems may actually execute. Explore the five trust domains behind DSG ONE.

Architectural model · Interactive 3D · No login required
SYSTEM / TOPOLOGY_01Five trust domains. One authority.
DSG // N2N
DRAG TO ORBIT · TAP TO INSPECT
CAM / ORBIT_PERSPECTIVE
FIG 01 / DSG_ONE
01 / SYSTEM ARCHITECTURE

Every layer has a boundary.

Tap a domain or select a node in the 3D model. The description stays readable even when animation is disabled.

DOMAIN 01 / GENERATE

Proposal / Agent Brain

User goals and model agents propose a bounded plan or candidate. The proposal may be optimized or rejected, but the proposer never grants itself execution permission.

PROPOSAL ONLY · NO SIDE EFFECTS
02 / OPERATING SEQUENCE

From intent to auditable result.

Each transition has an explicit contract. The route is conceptual—not a claim that any particular live E2E run passed.

01 →

Propose

Capture intent, constraints and candidate plan. Models generate options—not authority.

02 →

Verify

Check applicable deterministic constraints and bounded Z3 proof obligations.

03 →

Authorize

Spacetime evaluates plan, principal, scope, policy and approval requirements.

04 →

Execute

Only permitted steps reach governed AWS API, MCP, browser or tool adapters.

05 ↺

Record

Attach result, evidence, hashes and replay context. Incomplete proof is not success.

03 / DECISION POLICY

Three outcomes. No silent bypass.

DSG may continue authorized work. Missing permission pauses an in-plan step; unsupported requests fail closed.

ALLOW

Within the approved plan. The scoped step meets required authority, capability, constraints and evidence conditions.

WAITING_PERMISSION

Still within the plan. A required permission or capability has not been provisioned. Resume after authorization.

BLOCK

Out-of-plan or unsupported. Policy violations, missing authority or failed proof cannot be converted to a success claim.

04 / VERIFIED PRODUCTION EVIDENCE

Evidence over animation.

Verified 11 October 2026: AWS Spacetime source 5c09c3a68bebff32a543e543237cdd1b77a6b53c was activated with ECR image sha256:37c21d7947b624c742c63bafeb601c5ba8765021d788ca4aefd6d5b7eb0bc1c4. AWS activation run 38140649005 and signed read-only canary run 38140811436 passed; four peer containers were unchanged.

Scoped signed provider E2E: PASS. A real end-user OAuth principal executed the governed GitHub metadata read-only route. Provider/result and signed execution hashes, AWS SSM pinned public key, Ed25519 signature and tamper denial were independently checked. A fresh authenticated Production MCP readback on 11 Oct reported valid=true, records=140. This dated evidence does not prove Cinema remote Connect/Execute, Quest/Unity or whole-system E2E.

Read public signed E2E evidence ↗
AWS Spacetime deploymentVERIFIED · 11 OCT 2026
Production container healthVERIFIED · HEALTHY
HTTPS transport / anonymous denyVERIFIED · 401 FAIL-CLOSED
Evidence chain · observed 11 OctVALID · 140 RECORDS
OAuth → GitHub read-only → signed Ed25519 E2EVERIFIED · SCOPED PASS
Cinema remote Connect/Execute; Quest/UnityNOT VERIFIED
Spacetime Workroom storeUNAVAILABLE
Full-system Production E2ENOT VERIFIED
Source: real GitHub Actions activation logs #38140649005 and #38140811436, independently verified scoped read-only provider result and signature, and authenticated Spacetime MCP evidence-chain readback (11 Oct; 140 records). Evidence-chain count is a dated snapshot, not Cinema proof or live telemetry.
05 / THE NEXT STEP

Build with a boundary.
Verify with evidence.

For API integrations, governed execution or current release evidence, continue to the DSG ONE website or its public AWS status documentation.