On this page
Factor Agent
What It Solves#
Multi-agent system architecture using prime factorization as the decomposition principle. Agents are structured like factors of a number — each has a distinct identity, and compositions are computed through multiplication.
Key Innovation#
Prime Factorization as Architectural Decomposition:
Just as every integer factors uniquely into primes, every agent capability factors into prime-weighted primitives. This guarantees:
- Unique identity: No two agents have the same prime factor signature
- Composable authority: Agent capabilities combine through multiplication, not addition
- Invariant preservation: The Fundamental Theorem of Arithmetic ensures no authority is lost or duplicated
OTP Supervision as Computational Factorization#
The Elixir/OTP supervision tree implements:
- Factor decomposition: Processes are organized hierarchically by prime weight
- Restart strategies: Failed factors are replaced without disrupting the whole system
- Message passing: Inter-agent communication respects factor boundaries
The Repair Mechanism#
When a perturbation arrives, the system computes:
This is identical to the CRT repair in Geometry of Conditional Logic! The same safe-corridor mathematics applies to multi-agent repair.
Architecture#
TransactionalHarness
├── Compositional Invariants
├── Prime Factor Decomposition
└── Region-Based Repair (z' = z + kM)
Agent Primitives
├── p₁: Authentication
├── p₂: Authorization
├── p₃: Computation
└── pₙ: Custom capabilities
For the full distributed runtime architecture, including Elixir/OTP process trees, Horde recovery, ScyllaDB temporal state, lag-aware probes, and FactorAgent governance, see Agentic AI as a Distributed System.
Website#
Live: sethuiyer.github.io/factor-agent
Connection to Core Vision#
Factor Agent extends the Arithmetic Manifold to multi-agent systems. Prime factorization provides:
- A principled way to decompose agent capabilities
- Guaranteed compositionality (multiplicative, not additive)
- The same CRT-based repair mechanism as Geometry of Conditional Logic
The “factorization” of authority mirrors the “factorization” of constraints in the multiplicative PINN framework — both use the Euler product structure to ensure compositional correctness.
See Also#
- All Projects — project overview
- Agentic AI as a Distributed System — full Factor Agent runtime architecture
- Geometry of Conditional Logic — same CRT repair mechanism
- Authorization Lattice — capability composition via prime factorization
- Multiplicative PINN — Euler product structure in constraint enforcement
- Prime Weighting — how primes enable factor decomposition
- The Arithmetic Manifold — the unified theory