Method
Biology as hypothesis,never as evidence.
Mutuus treats every biological analog as a source of engineering hypotheses. A candidate advances only by passing explicit gates, and it is measured on its lifecycle economics rather than a cold-start microbenchmark.
Inorganic
Fixed internal organization, fixed tradeoffs, no explicit lifecycle. A Vec at allocation time and after ten billion operations is governed by the same model.
Organic
Lifecycle state that evolves through workload exposure and maintenance, moving through thermal states such as Hot → Warm → Cold → Compressed.
Efficiency is the baseline,not the whole story.
O(e)
Efficiency
Classical time and space bounds for every operation, published beside the inorganic counterpart. The baseline: an organic dramatically slower at its core operations is not viable.
O(a)
Adaptiveness
Convergence speed, adaptation quality, and catastrophic forgetting. Inorganics have O(a) = 0 by definition.
O(r)
Resilience
Behavior under adversarial input, capacity overflow, and skipped maintenance. Organics should degrade smoothly toward inorganic baseline behavior.
O(τ)
Thermal cost
Per-tick overhead, amortized self-management cost, and the break-even point. The most important dimension, and the most neglected.
Most candidateswill not survive.
We expect to abandon more candidates than we publish, and consider that healthy. A candidate that fails a gate is revised or abandoned, not advanced on rhetorical appeal.
- 01
Nature analog identification
Does the biology give real computational insight, or is it a naming exercise? Convergent analogs across at least two phyla are required.
- 02
Formal specification
Data model, operations, complexity bounds, and invariants, precise enough to implement from.
- 03
Complexity analysis
Theoretical bounds for all operations against the inorganic counterpart. It must win at least one operation class.
- 04
Implementation
Working Rust with comprehensive test coverage that matches the specification.
- 05
Benchmarking
Reproducible Criterion.rs benchmarks across workloads, sizes, and access patterns. Wins, losses, regressions, and dead ends are published.
- 06
WASM + TypeScript
Bindings that make the organic usable beyond Rust, with acceptable overhead in the browser.
- 07
Publication & artifacts
Papers, results, and interactive simulations. Not complete until others can reproduce and build on it.
“Every organic has a metabolic cost, and the response is not to eliminate the metabolism but to let the organism grow compensatory structures through its lifecycle.”
Step 1
Construct both structures with identical data.
Step 2
Exercise both across at least 100 tick cycles.
Step 3
Then benchmark, once adaptation has developed.
Step 4
Report t = 0 and steady-state results, together.
Not a rejection of classical computer science.
The four dimensions extend O(n) analysis; they do not replace it.
Not biologically faithful simulation.
The analog is a reasoning tool. Implementations are judged on their own merits.
Not “everything should be organic.”
Many workloads are best served by simple, predictable inorganic structures.
Not performance marketing.
The primitive papers publish losses alongside wins.
Source
Adaptive Computing Primitives: A Position Paper on Bio-Inspired Structures, Policies, and Coordination
B. D. Phillips · Mutuus Research · 2026