MD-OS (Artificial Prefrontal Cortex) v5.0 separates operational reproducibility from epistemic truth.
The filesystem can make a reasoning path readable, auditable, reconstructible, and actionable. It does not make a theory true by itself. Every theoretical or computational result should therefore carry an epistemic status that says what kind of claim is being made.
MD-OS can be described as a linearizer of thought in geometric-procedural form.
This phrase is useful only if it stays operational. It does not mean that thought is reduced to a single flat sentence. It means that fluid or heuristic cognition is externalized into a navigable operational space and then routed through bounded procedures:
fluid or heuristic thought
-> textual nodes
-> relations and dependencies
-> procedures
-> states
-> checks
-> reconstruction and replay
The model is geometric because ideas, clues, results, and assumptions become a space of files, links, hierarchies, indices, graphs, and operational boundaries.
The model is procedural because the same material is not left as contemplation. It becomes actions, builders, gates, audits, states, and transitions.
In short:
intuition -> trajectory
trajectory -> structure
structure -> procedure
procedure -> verifiable memory
The practical outcome is claim discipline. MD-OS should make it clear whether a result is still a clue, a line of thought, a source principle, a reproducible derivation, a prediction/readback artifact, or a correction.
heuristic material
-> line of thought
-> frozen principle
-> derivation
-> prediction or readback
-> correction
-> replay
In MD-OS terms:
heuristic: notes, journal entries, intake material, and raw analysis.line_of_thought: note blocks, natural-language programs, work items, and
dependency sketches that organize a possible path.frozen_principle: source-of-truth files such as md-os/kb/**, ME.md,
AGENTS.md, and canonical programs.derivation: deterministic scripts, symbolic calculations, builders, and
reproducible transforms.prediction_or_readback: research results, connector outputs, gates,
audits, and comparison artifacts.correction: change proposals, source edits, rebuilds, and replay.| Phase | Typical files or artifacts | Promotion condition | Failure to avoid |
|---|---|---|---|
heuristic |
notes, journal entries, intake material, raw analysis | The clue is organized into an explicit relation or work item. | Treating a useful intuition as a claim. |
line_of_thought |
linked notes, natural-language programs, work items, dependency sketches | Assumptions and dependencies are clear enough to freeze or test. | Hiding missing assumptions inside fluent prose. |
frozen_principle |
md-os/kb/**, ME.md, AGENTS.md, canonical programs |
The principle is accepted as source of truth for future operation. | Freezing a principle without reviewable scope and limits. |
derivation |
deterministic scripts, builders, reproducible transforms | Inputs, method, and outputs are recorded and repeatable. | Presenting a derived result without declared assumptions. |
prediction_or_readback |
gates, connector snapshots, audit outputs, comparison artifacts | The comparison target and method are explicit. | Calling retrodiction a strict prediction. |
correction |
change proposals, source edits, rebuilds, replay reports | The correction changes source or generated state with visible evidence. | Correcting silently without preserving why the claim changed. |
A claim may be promoted only when its previous phase, assumptions, evidence, and failure condition are visible.
Use this minimum checklist before promoting theory-oriented material:
What phase is it in now?
What assumptions does it depend on?
What source or artifact records it?
What would count as a failed gate?
Was the target declared before readback?
What changed if this is a correction?
If those answers are missing, the result should remain heuristic,
conditional, derived, or open.
In this model, logic as logos is the ordered chain of ideas. It is broader than formal calculation alone: it is thought arranged as a coherent trajectory with explicit constraints.
Clues are local segments of that trajectory:
clue
-> relation
-> thought segment
-> coherent line
-> principle
-> derivation
-> prediction
MD-OS should not leave these segments scattered. It turns them into an operational geometry:
The threshold is explicit:
suggestive chain = heuristic
constrained chain = operational logos
constrained chain + calculation + failure risk = predictive theory
Computational and theoretical artifacts should use one of:
heuristic
conditional
derived
retrodictive
predictive
open
falsified
The status belongs to MD-OS, not to the calculation engine. A calculator may
return True, but MD-OS still has to say whether the result is derived under
declared assumptions, conditional on an unfinished proof, retrodictive against
known data, predictive before readback, open, or falsified.
Retrodiction must not be presented as strict prediction.
A result is predictive only when the prediction target, timestamp or version,
input state, assumptions, and comparison procedure were declared before
readback. Otherwise it is retrodictive, conditional, derived, or
open.
Every theory-oriented work item, calculation profile, gate, or research result should therefore state which phase it occupies:
heuristic
line_of_thought
frozen_principle
derivation
prediction_or_readback
correction
The same artifact may move through phases over time, but the movement must be visible through source edits, change proposals, generated state, journal events, rebuilds, or replay. MD-OS may stabilize and audit the path, but it does not make the theory true merely by storing it.