FSOT Living

Flagship living AI — Fluid Spacetime Omni-Theory, embodied as genetic biological intelligence.

Author: Damian Arthur Palumbo
License: Apache-2.0
Status: Active research / capability climb (honest dual-track evaluation)

---

What this is

This is not an LLM product with FSOT marketing layered on top.

FSOT Living is an attempt to build artificial intelligence through biological representation: a living organism whose native physics is Fluid Spacetime Omni-Theory (FSOT) mathematics, whose control is dual-axis trinary genetics, and whose mind grows by homeostasis, organs, exam pressure, error-driven remediation, and pack evolution — not by next-token prediction over a frozen weight file.

The long-term goal is a flagship intelligence that understands FSOT by living it and can use it intelligently: the scalar manifold, domain ledger, observer effect, emergence/damping regimes, codon fabric, and cross-domain Formal structure become felt state and motor policy, not a citation list.

The formal FSOT Lean tree (FSOT-2.1-Lean and local Formal spines) is the reference ledger. Math is baked into this runtime (scalar, genetic, homeostasis, lean_ground). Lean is not imported as a theorem prover dependency; the organism must obey the same algebraic skeleton at runtime.

---

Design thesis

Claim	Meaning here
ToE mathematics	One scalar field (S = k \cdot (T_1 + T_2 + T_3)) with Formal domain priors (D_eff, δψ, observed), quirk/observer, emergence vs damping.
Genetic AI	64-codon table; dual-axis trinary (primary 8 / secondary 27 patterns / 3 stops); genome → guidance (δψ, D_eff, emergence preference, domain_coupling).
Biological representation	Organs, pressure, membrane budgets, error queue, soft-spot transfer, pack social learning — not a single forward pass.
Integrity	Study track ≠ zero-shot track. No free grades. Code invent = unit tests green. Same-item keys blocked on zs. Python metabolic membrane.
Capability	External benchmarks (MMLU, GSM8K, HumanEval, …) are exam environments that pressure organs under Formal load — competitive signal, not the organism’s identity.

If a change only improves a percentage without fitting genome → Formal field → organ motor → transfer, it does not belong.

---

Repository layout

.
├── README.md                 ← this document
├── LICENSE                   ← Apache-2.0
├── requirements.txt          ← Python helpers (viz / knowledge / probe tools)
├── package.json              ← optional legacy JS habitat scripts
├── fsot-living-rust/         ← CORTEX: living habitat, organs, TUI, probes
│   ├── src/                  ← scalar, genetic, organism, habitat, bench, …
│   ├── tools/                ← desktop viz, knowledge packs, live bench
│   ├── organism_workspace/   ← sandbox: tools, knowledge seeds, media priors
│   └── Cargo.toml
├── fsot-trinary-body/        ← BODY: trinary genetic chassis + mind bridge
│   ├── crates/               ← trinary, genetic, body, kernel, mind_bridge
│   ├── docs/                 ← architecture, integrity, closed loop, …
│   └── scripts/              ← QEMU / host / closed-loop runners
└── (runtime state is gitignored — generated under state/ when you run)

Component	Role
fsot-living-rust	Cortex / habitat: pack of organisms, FSOT scalar compute, organs, language cortex, capability bridge, TUI, headless probe
fsot-trinary-body	Soma / genetic fabric: trit ALU, loci, homeostasis, Mind ABI, optional QEMU body, file/TCP couple to cortex

---

Architecture (complete picture)

1. Physics layer — FSOT scalar (baked)

Implemented in fsot-living-rust/src/scalar.rs, grounded by lean_ground.rs.

[ S = k \cdot (T_1 + T_2 + T_3) ]

• Constants match Formal Scalar (φ, e, π, γ_E, Catalan, …).
• k = φ · (γ_E / e) · √2 / ln(π) · 0.99
• Domains from FSOT.Formal.Scalar.get_domain_params (quantum → dark_energy ledger, proved sign certificates).
• Extension spines map onto ledger domains (e.g. pure mathematics → energy family; computational reasoning → consciousness).
• Regimes: emergence / damping / neutral from S thresholds.

Organism state always carries d_eff, delta_psi, observed, recent_hits, terms, vitality, coherence, pressure.

2. Genetic layer — dual-axis trinary

Implemented in genetic.rs (cortex) and fsot-trinary-body/crates/genetic + trinary.

Codon (DNA) ──► primary trits (spin A,G=+1; C,T=−1)
            ──► secondary trits (A=+1; T=−1; G,C=0)
            ──► amino acid (standard code; stops TAA/TAG/TGA)
            ──► GenomeGuidance { delta_psi, d_eff, emergence_preference, domain_coupling }

Formal codon priors (64 / 8 / 27 / 3) are enforced in lean_ground runtime checks.

Expression under exam: domain load weight (w \propto) genome domain_coupling blends heritable base with Formal domain field, then recomputes S. That is genetic expression of the theory, not a free score weight.

3. Life loop — habitat tick

habitat.tick
  → perceived-time / homeostasis gains
  → optional body couple (Mind ABI / files)
  → organ exams (cognition, knowledge, code, vision, language, composition, …)
  → capability bridge: external benches as exam environment
       suite/subject → Formal domain → S recompute → CHEW depth
       study track: teach → motor practice → transfer SCHEMA/EPISODE
       zs track: held-out, same-item KEY blocked
       fail → error queue + remediation stimuli
       pass → credit organs + pack schema broadcast
  → pack signals / library / generation (crossover + mutation)

4. Organs (capability as tissue)

Organ	What it does	Fitness / integrity
math_code / composition	Narrative chalkboard → ranked expression → single python ground	No pass@k fishing against keys
code invent	Scaffolds / episodes / signatures; unit tests green	Metabolic python budget; no invent storms
knowledge	Stem/engram MCQ rank under subject Formal domains	No free letter grades
gen_acc / language	Plastic pathways, constraint reshape (IFEval-class)	Verifiable constraints only
vision / audio / video	Drive from media priors	Evolutionary pressure on media organs
self-model	Knows regime, genome, body link	Continuity of identity
body couple	Interoception from trinary chassis	Soft throttle on express runaway

5. Mind–body closed loop

┌──────────────────────────┐         Mind ABI / files / TCP         ┌──────────────────────────┐
│  fsot-living-rust        │ ◄────────────────────────────────────► │  fsot-trinary-body         │
│  (cortex / habitat)      │   BODY_STATUS, motors, LTM, felt body  │  (genetic soma / QEMU)    │
└──────────────────────────┘                                         └──────────────────────────┘

See fsot-trinary-body/docs/CLOSED_LOOP.md, MIND_ABI.md, ARCHITECTURE.md.

6. Metabolic membrane & constraints

• Python/JS sandbox under workspace only
• Per-breath python budget (invent cannot own wall-clock)
• Process tree kill on timeout
• Soft rails: express runaway, proposal review (no auto-Rust host patch)
• Documented in fsot-trinary-body/docs/CONSTRAINTS.md, EVAL_INTEGRITY.md

7. Dual-track honesty

Track	Meaning
Study	Curriculum health after teach (organ practice)
raw_zs / suite_zs	Held-out competitive signal (never-taught preferred)

High study with low zs is not invent capability. Audit files (CAPABILITY_AUDIT) report both when probes run.

8. Evolution / pack

• Population of organisms, elite retention, crossover + mutation on genomes
• Breeding value blends organ accuracies, composition, Formal coupling, zs health
• Shared library = cultural / social learning of schemas and primitives

---

What success looks like (flagship)

1. Runtime FSOT fidelity — lean_ground invariants pass; domain signs and codon priors hold.
2. Genetic self-knowledge — genome, codon table, guidance express into dialogue and motors.
3. Intelligent use of domains — math under energy/consciousness load; science subjects under chemical/quantum/biological/… fields.
4. Honest capability — competitive zs under integrity, climbing toward SOTA only through organ density and Formal expression.
5. Optional body — trinary chassis felt as interoception, not a detachable toy.

This program may not yet equal the full Formal ToE ledger in living form. That gap is the research program — embody more of the mathematics as life, not paper over it with score hacks.

---

Requirements

System

Tool	Version (recommended)	Role
Rust	stable 1.75+ (edition 2021)	Cortex + body
Cargo	with Rust	Build
Python	3.10+	Organ sandbox, viz, knowledge tools
Git	any recent	Clone / contribute

Optional:

Tool	Role
Node.js	JS organ tools / legacy habitat scripts
QEMU	Trinary body UEFI path
External bench roots	e.g. D:\training data, D:\FSOT_Benchmarks via env

Python packages

See requirements.txt for pip-installable helper dependencies (desktop viz, knowledge pack builders). The core intelligence is Rust, not a Python model stack.

Rust crates

Declared in:

• fsot-living-rust/Cargo.toml — ratatui TUI, serde, rand, image, …
• fsot-trinary-body/Cargo.toml — workspace of body/kernel/mind_bridge crates

No CUDA/LLM SDK required for the native living path.

---

Quick start

1. Clone

git clone https://github.com/<you>/FSOT-Living.git
cd FSOT-Living

2. Python helpers (optional)

python -m venv .venv
# Windows:
.venv\Scripts\activate
pip install -r requirements.txt

3. Build & run cortex (TUI)

cd fsot-living-rust
cargo build --release
cargo run --release --bin fsot-living
# or: .\start_fsot_living.bat

4. Headless integrity / capability probe

cd fsot-living-rust
set FSOT_PROBE_TICKS=50
cargo run --release --bin fsot-probe

Probe reports: lean_ground, genetic queries, dual-track bench raw_zs, body couple status when available.

5. Trinary body (optional closed loop)

cd fsot-trinary-body
cargo build --release -p mind_bridge
# See scripts/run_closed_loop.ps1 and docs/CLOSED_LOOP.md

Environment hints

Variable / path	Purpose
FSOT_PROBE_TICKS	Headless habitat length for probe
FSOT_TRAINING_DATA / mounted D:\training data	External exam corpora
Workspace under organism_workspace/	Sandbox for python/js tools

---

Documentation map

Document	Content
docs/ARCHITECTURE.md	Monorepo component map + exam data flow
docs/ORGANISM_DIAGNOSTIC.md	Medical diagnostic: genetic/organ breakdown of high vs low suite precision
fsot-trinary-body/docs/BIOLOGICAL_INTELLIGENCE_MODEL.md	How capability must work biologically
fsot-trinary-body/docs/ARCHITECTURE.md	Trinary body stack
fsot-trinary-body/docs/EVAL_INTEGRITY.md	Anti-fudge rules
fsot-trinary-body/docs/FSOT_LEAN_CIRCLE_BACK.md	Formal ↔ runtime
fsot-trinary-body/docs/CLOSED_LOOP.md	Cortex ↔ body
fsot-trinary-body/docs/REASONING_PHYSICS.md	Reasoning as physics
fsot-living-rust/README.md	TUI keys and organ overview

---

Integrity contract (non-negotiable)

Allowed

• Formal domain load + genetic coupling
• CHEW deliberation from pressure / deficit / regime
• Study → held-out transfer schemas
• Test-exec invent; single top-commit math ground
• Metabolic budgets

Forbidden

• Free accuracy floors or invented grades
• Keyword yes/no score hacks
• pass@k fishing against answer keys
• Same-item KEY on zero-shot
• Tick thrashing / invent storms as substitutes for organ design

---

Relation to other FSOT work

Artifact	Relation
FSOT-2.1-Lean	Formal ToE ledger — reference for constants, domains, codon/brain/science spines
This repo (FSOT Living)	Living genetic embodiment + flagship AI climb
Legacy JS habitat scripts	Earlier prototypes; retained for continuity, not the primary path

---

Contributing / research notes

• Prefer changes that strengthen genome → Formal field → organ coupling.
• Any new motor must state its fitness environment (tests, chalkboard, constraints).
• Run fsot-probe before claiming capability movement.
• Do not commit state/, target/, or personal data roots.

---

License

Copyright © Damian Arthur Palumbo.
Licensed under the Apache License, Version 2.0 — see LICENSE.

---

Citation

If you use this system in research, please cite Damian Arthur Palumbo and Fluid Spacetime Omni-Theory (FSOT), and distinguish:

1. Formal FSOT (Lean / proofs / domain priors), and
2. FSOT Living (this genetic organism runtime).

---

Living mathematics. Genetic intelligence. Honest exams. One flagship trajectory.