FSOT Mathematica A

Fluid Spacetime Omni-Theory as a Mathematica formula microscope, living observer mind, and FSOT-native language path — with Python parity for train/generate.

License: Apache License 2.0
Repo: https://github.com/dappalumbo91/FSOT-Mathematica-A

Local clone path (optional): I:\fsot in mathmatica
Verification authority (separate archive): Lean / multi-prover hub + GPU seed triangulation

What this is

1. Readable mathematics — seeds → constants → S = K(T1+T2+T3)
2. Living observer mind — STM/LTM, domain routing, think/dream
3. FSOT language model — embed, collapse, consensus attention (no softmax), suction–poof train, generate, math trace ledger
4. Conversation + explain — multi-turn dialogue with claim-bank substance (not empty occupation rhetoric)

Zero free parameters beyond seeds {π, e, φ, γ, G} on the FSOT spine. Every operator is named and traceable.

Scientific continuity

Prior literature still teaches its regimes (Newtonian mechanics, relativity, Standard Model, …). FSOT unifies under a seed spine; it does not erase the path that led here. See docs/SCIENTIFIC_CONTINUITY.md.

---

Clone

git clone https://github.com/dappalumbo91/FSOT-Mathematica-A.git
cd FSOT-Mathematica-A

Quick start

Mathematica / Wolfram

Adjust the path to your clone:

Get["/path/to/FSOT-Mathematica-A/FSOT/init.wl"]

(* Authority parity with shipped seed JSON *)
FSOTLLMVerifyAuthority[]

(* Language model *)
FSOTLLMNew[]
cur = Import["/path/to/FSOT-Mathematica-A/data/fsot_llm_curriculum.json", "RawJSON"]["pairs"]
FSOTLLMTrain[cur, 5]
FSOTLLMGenerate["fluid spacetime communicate language", 16]
FSOTLLMLastTrace[]
FSOTLLMFormulaSheet[]

(* Living mind *)
FSOTAwaken[]
FSOTThink["Universal fluid communication across domains."]

(* Formula twin *)
FSOTShowFormulas[]

(* Microscope after a Python export *)
FSOTMicroscopeLoad[]
FSOTMicroscopePlotParts[1]
FSOTMicroscopeStructured[]
FSOTMicroscopeParagraphs[]
FSOTMicroscopeConversation[]

Why Mathematica: docs/WHY_MATHEMATICA.md — formula microscope (not a Lean replacement). Python trains/runs without Wolfram Engine.

Python (train / conversation)

cd FSOT-Mathematica-A
python scripts/run_fsot_llm_python.py          # full train + microscope export
python scripts/fsot_conversation.py --smoke   # multi-turn demo
python scripts/fsot_conversation.py --chat    # interactive
python scripts/fsot_explain.py --once "Does FSOT replace relativity?" --deep

PFLT bridge (refresh slot prefers, no full retrain)

Optional: if you have Desktop PFLT data paths configured in the scripts.

python scripts/refresh_pflt_bridge.py

Ontology note: ad-hoc free params steer when direction is unknown; FSOT fixes direction via seeds + domain occupation — see docs/AD_HOC_VS_ZERO_FREE.md.

Conversation notes

See docs/CONVERSATION.md. Explain mode uses data/explanatory_pack.json (claims first). Deep arcs use 6–8 step paragraph tissue when not answering how/why.

python scripts/expand_fluency_curriculum.py --quick-train
python scripts/fsot_conversation.py --chat

Wolfram Engine (optional):

wolframscript -file "examples/FSOT_LLM_Train_Generate.wls"

---

Layout

FSOT-Mathematica-A/
  FSOT/           Scalar, domains, shared embeds, init
  llm/            FSOTLLM.wl — LLM engine + traces/
  living/         Living observer mind
  memory/         Persisted mind + model JSON
  data/           Seeds authority, curriculum, explanatory pack
  scripts/        Python train, conversation, explain, PFLT bridge
  examples/       Train/generate + living awaken scripts
  docs/           Continuity, conversation, allocation, LLM design
  bridge/         Links to related FSOT projects
  LICENSE         Apache-2.0
  NOTICE

---

How the pieces fit

System	What FSOT does there
Lean archive	Multi-domain verification ledger
PFLT	Proto-Fluid Language Translator — linguistics and universal communication (full language surface is the goal)
FSOT-GPU	Collapse, consensus attention, CUDA, industry LM host
This Mathematica home	Same math, visible operators + LLM train/generate + living mind

PFLT and this LLM path are aligned: domain folds, scalar modulation, no ad-hoc free-parameter core. Mathematica is where you see every relationship in the ledger.

---

Verification

• data/fsot_seeds_authority.json — from GPU/archive kernel constants
• FSOTLLMVerifyAuthority[] — live WL vs authority
• Traces: llm/traces/last_forward_trace.json

---

Stance

Gaps in logic get filled. Applications are real work across domains. Recognition is an uphill battle; the stack is built to be checkable (GitHub, X, verification bundle) so finding it is a matter of time, not of diluting the theory.