Constraint-First Systems Architect
London, UK
I design deterministic control primitives for AI systems.
My work focuses on pre-execution governance: explicit authority, halt-first design, and minimising degrees of freedom before optimisation is introduced.
This profile contains small, auditable public primitives.
Composition and orchestration logic remain private by design.
Governance begins before execution.
Most AI systems optimise first and constrain later.
I design systems where:
- Constraints are explicit
- Execution requires authority
- Halt is a structural capability
- Behaviour is deterministic
- Failure modes are defined in advance
No narrative compliance.
No hidden autonomy.
No optimisation-first architecture.
The repositories below form a coherent control layer stack.
Finite-state stop controller (GREEN → AMBER → RED).
RED is absorbing. Deterministic transitions. Fully tested.
Purpose: Make "halt" a mechanical property of the system rather than a policy suggestion.
Hash-based invariant locking for configuration and execution boundaries.
Purpose: Prevent silent drift in systems that must remain stable over time.
Public workbench of small deterministic control primitives.
Purpose: Publish the bricks, not the aircraft blueprint.
Typed, versioned language primitives for reducing ambiguity in AI governance contexts.
Purpose: Reduce interpretive drift by constraining vocabulary.
Static analysis for detecting ambiguity, missing halt semantics, and weak constraint definitions in policy text.
Purpose: Make policy mechanically inspectable.
Experimental boundary-testing primitives exploring admissibility and explicit authority gates.
Purpose: Stress-test execution assumptions before deployment.
- Determinism over optimisation
- Explicit authority required for execution
- Stop is a first-class primitive
- Shrink degrees of freedom before adding complexity
- Tests are mandatory
- Public artefacts do not expose private orchestration
📄 Zenodo:
https://zenodo.org/search?q=ricky%20dean%20jones
🔗 LinkedIn:
https://linkedin.com/in/ricky-jones-1b745474
- Pre-execution admissibility layers
- Halt-first AI architecture
- Deterministic control in large model environments
- Structural governance beyond narrative compliance
Small, auditable tools.
Clear failure modes.
Minimal claims.
Running code over commentary.