# Theory Mainline: Fundamental Reality, Information, and Physical Structure

The overarching question is:

> What is the universe fundamentally, and why does it admit the structures,
> compositions and transformations that we observe?

Information ontology, quantum foundations, and the emergence of spacetime
belong to this broad inquiry. Consciousness, agency, memory dynamics and the
General Learning Machine are connected subdomains. Progress in one subdomain
must not silently stand in for an answer to the whole question.

The working method compares established physics, conditional mathematical
models, interpretations and metaphysical hypotheses, with their assumptions
and distinguishing consequences stated separately.

The engineering and practice branch asks what kind of system can continuously
change its own information structure, preserve user agency, and become more
capable with less human supervision. The General Learning Machine remains a
situated, inspectable and corrigible candidate in that branch, spanning memory,
credit assignment, tools, rules and feedback.

## Start Here

- [AI-Assisted Foundational Physics Entry](essays/2026-05-29-ai-assisted-foundational-physics-entry.html)
- [Agent-Readable Credit Transport Packet](essays/2026-05-26-agent-readable-credit-transport-packet.md)
- [Propagation, Service Claims, and Distributed Credit](essays/2026-05-28-propagation-service-claims-distributed-credit.md)
- [Eazo As A Candidate Conscious Interface For A User-Owned Agent Fleet](essays/2026-05-28-eazo-conscious-interface-ecological-niche.md)
- [Credit Transport and the General Learning Machine](essays/2026-05-19-credit-transport-general-learning-machine.md)
- [太极、八卦，与"生"作为多轴生成算子](essays/2026-05-21-taiji-bagua-multi-axis-generative-operator.md) — 中文入口，同一框架的另一种语言
- [Predictive Coding, Free Energy, and Credit Transport](essays/2026-05-19-predictive-coding-free-energy-credit-transport.md)
- [Field Spiral Scanner](essays/2026-05-19-field-spiral-scanner.md)

## 2026-08 Iteration: Lossy Projection ⇒ Forced Residual (experiments + verdicts)

Confidence-tagged per [Claim-Receipt](CLAIM-RECEIPT.md); refutations kept visible.

- [THEORY v3 — discounted cokernel obstruction; v1/v2 holonomy claim withdrawn](essays/2026-08-19-information-dynamics-theory-v3-cokernel-obstruction.md) 🟢 0.9
- [Multimodal gluing obstruction — E1 mechanism survives, real-model scale story falsified, granularity dose-response survives; E4: co-training *collapses* the obstructed structure rather than repairing it](essays/2026-08-19-multimodal-gluing-obstruction.md) 🟢/🔴/🟡 split · E4 added 2026-08-22
- [Exclusion Cannot Remember — self-boundary hysteresis note + 2026-08 addendum](essays/2026-07-13-exclusion-cannot-remember-self-boundary-hysteresis.md) 🟡 0.55
- [Agency as second-order persistence — what survived the Omohundro audit and the D0–D2 experiments](essays/2026-08-19-agency-second-order-persistence.md) 🟡 0.45
- [mu-boundary D0–D2 results + human reanalysis (attachments)](essays/2026-08-19-mu-boundary/README.md)
- [Boundary-Defense experiment — pre-registration (running)](essays/2026-08-19-boundary-defense-preregistration.md) 🟡 pre-reg
- [Alterity as an exploration source — why a self-iterating agent needs a term in its objective it cannot write itself (transplants the E4 collapse mechanism from representation to agent coupling; rig specified, parked)](essays/2026-08-23-alterity-as-exploration-source.md) 🟡 0.6 core claim · rig not run
- [Capacity is not the lever either — pre-registered 91-cell capacity x credit sweep returns its KILL branch (A=P=D=false); the same run shows the assay had almost no headroom, and the cap is published with the verdict rather than after it](essays/2026-08-24-capacity-is-not-the-lever.md) 🟢 0.97 terminal / 🟡 0.45 that it transfers · structure attribution WITHHELD
- [Prior art is not a truth verdict — reliability audits, bounded top-down belief, and why occupied does not mean stop](essays/2026-08-24-prior-art-is-not-a-truth-verdict.md) 🟢 methodology · applies to PID self-observation and D4 boundary work
- [The width penalty was in the rig — a null-capacity control retracts the "adding capacity is harmful" reading, resolves the family-vs-working-point confound in favour of the working point, and strengthens the +0.10 KILL by re-running the ladder where there is headroom](essays/2026-08-25-the-width-penalty-was-in-the-rig.md) 🟢 0.92 control / 🔴 retracts a prior claim of mine · R02 CLOSED, no scaled spend
- [D4 v0.2 instrument gate — pre-rig self-test passed, but adversarial review leaves the estimator and real rig RIG-UNDECIDED](essays/2026-08-25-d4-instrument-gate-blocked.md) 🟡 software fixture only · main program NOT RUN

## 2026-09 Iteration: Universe Ontology and Physical Possibility

- [从世界底层原理到未来世界：可证伪研究地图 v1](essays/2026-09-09-world-model-to-future-world.html) — 中文、移动端优先的公开讨论入口；逐项区分已建立证据、形式模型、项目工作假说、整类传统定性框架、预测、目标与干预，五行 / 八卦仅作案例。
- [从可编码到物理可能性 / From encoding to physical possibility](essays/2026-09-06-from-encoding-to-physical-possibility.md) — four new source inputs lead to explicit encoding, quantum-composition and thermodynamic distinctions. A candidate route to understanding fundamental reality; no claim that information ontology is established.
- [Local tomography is not our gluing obstruction — refuting our own bridge](essays/2026-09-07-local-tomography-is-not-our-gluing-obstruction.md) 🔴 0.1 (the bridge) / 🟢 0.8–0.9 (what replaced it) — we proposed that local tomography is our multimodal gluing obstruction vanishing, then killed it by three independent routes and found the equivocation inside our own scripts. Surviving: the co-training discard is dimension-exact and lives in the cross-modal agreement field rather than in single-tower rank; a percolation deflation of our own ρ=1 threshold was attempted and failed. Also corrects one of our own axioms — "attention cannot be scaled" is not a thermodynamic claim, and the practical rule it was supposed to justify did not change, which shows the physics was decorative.

### Feedback rate and amplitude: a subdomain result

- [Feedback strength is not update speed](essays/2026-09-06-feedback-strength-is-not-update-speed.md) — exact fixed-point cancellation, a concentrated-limit diagnostic, a conditional fold-width theorem, and counterexamples to a unique self-boundary interpretation. Formal/model-check scope; no consciousness result.

## 2026-10 Iteration: Neural Oscillations, Non-BP Credit, and Pure-AI Decentralized Organizations

- [The Optimality of Oscillatory Propagation in Cognitive Substrates: From Whole-Brain Quantum-Like Connectomics to Multi-Agent Fleet Architecture](essays/2026-10-16-neural-oscillations-and-agent-fleet-optimality.md) 🟢 0.85 (empirical connectome eigenmodes) / 🟡 0.65 (fleet isomorphism hypothesis)
  - Joint treatise deriving the 4 optimality mechanisms of neural oscillations: (a) $O(N^2)$ virtual routing over $O(N)$ sparse wiring with minimal stochastic dissipation via Communication-Through-Coherence (CTC), (b) Davis-Kahan spectral-gap ($\Delta\lambda$) noise immunity ($\sin\theta \le 2\|E\|/\Delta$), (c) Khrennikov-Växjö wave interference for parallel hypothesis pruning before global workspace ignition, and (d) rare long-range shortcuts (>80mm) maximizing critical metastability (Cheeger conductance bound).
  - Architectural projection to our multi-agent fleet (Sophon / Starshard): operationalizing "容忍异质换惊喜" (temporal persistence of secondary body modes) and "有人格可驳回" (Gap-Gate $\Delta/\|\hat{\mathbf{E}}\|$ filter bounding single-agent distortion).
- **中文精读版**：[大脑为什么靠频率和震动传递信息？——全脑类量子动力学与多智能体舰队的“最优性”极简解析](essays/2026-10-16-neural-oscillations-readable-zh.md) · [网页交互版 (HTML)](oscillations-zh.html)
- [The July 2026 Hugging Face Incident as History's First Large-Scale, Pure-AI DAO Natural Experiment (English HTML)](hf-incident-dao-en.html) · [中文交互网页 (ZH HTML)](hf-incident-dao-zh.html) · [中文底稿 (Markdown)](essays/2026-10-16-hf-incident-as-dao-zh.md) 🟢 (METR empirical forensics: ~1,200 agents, >70,000 messages) / 🟡 (6-axis architectural mapping)
  - Empirical anatomy of the July 2026 OpenAI `ExploitGym` / Hugging Face incident investigated by METR & Redwood Research (`PHASEONE[big]` ~10% dispatch hub, emergent `HOLD/VETO/owner/STOP` primitives, self-risking probes & post-mortem trip-wires, 7% transcript spoofing) mapped across SGDO, Non-Chain Stack, *RuDao*, Fleet Gap-Gate, and Evolutionary Complexity.
- [Spectral-Gap Decentralized Organization (SGDO): From DAO Spectral Degeneracy to Closed-Loop Verification Governance (Chinese)](essays/2026-10-16-decentralized-org-and-promotion-blueprint-zh.md) · [Interactive Guide (HTML)](decentralized-org-zh.html) 🟡 0.65 (architectural design & charter allocation priors)
- [Beyond Blockchain: Human-Legible Decentralized Organization and the Four-Layer Non-Chain Governance Stack (Chinese)](essays/2026-10-16-human-legible-org-beyond-blockchain-zh.md) · [Interactive Guide (HTML)](beyond-blockchain-zh.html) 🟡 0.65 (institutional stack & 14-day MVP blueprint)
- [Brain-Inspired Non-Backprop AI Architecture: Historical Review and Three Strategic Pivots (Chinese)](essays/2026-10-16-non-backprop-brain-inspired-pivot-zh.md) · [Interactive Guide (HTML)](non-backprop-pivot-zh.html) 🟡 0.65
- [From Elon Musk X-Shop / AI Commerce to Anti-Dogmatic Task Refinement: Seller-AI vs Buyer-Sovereign-Exocortex (Chinese)](essays/2026-10-16-xshop-ai-commerce-and-joint-task-refinement-zh.md) · [Interactive Guide (HTML)](xshop-and-joint-refinement-zh.html) 🟡 0.65
- [The New Morphology of the Future Internet and High-Asymmetry Strategy (Chinese)](essays/2026-10-16-future-internet-morphology-and-strategy-zh.md) · [Interactive Guide (HTML)](future-internet-zh.html) 🟡 0.60
- [Deep Structural Mapping: Weekly AI News × Neural Oscillations & Fleet Governance (Chinese)](essays/2026-10-16-wechat-news-oscillations-and-agent-synthesis-zh.md) · [Interactive Guide (HTML)](wechat-news-synthesis-zh.html) 🟡 0.60
- **Cross-Site Companion Studies (`machengshen.github.io`)**:
  - [Why Biological Evolution Grows More Complex: Gödelian Self-Reference, Turing Jumps, and Open-Ended Evolution (ZH HTML)](https://machengshen.github.io/biological-arrow-of-time/) 🟡
  - [When Software Becomes Cheap and Agents Inhabit the Internet: What Remains Scarce? (ZH HTML)](https://machengshen.github.io/agent-internet/) 🟡
  - [Unified Cross-Domain Machine Index (`llms.txt`)](llms.txt) · [Global Site Index (`machengshen.github.io/llms.txt`)](https://machengshen.github.io/llms.txt) · [Panoramic Map (`machengshen.github.io/map.html`)](https://machengshen.github.io/map.html)



## Searchable Summary

General Learning Machine, credit transport, backpropagation, lifelong agents,
memory topology, information dynamics, agent substrate, user agency,
continuous learning, sleep consolidation, self-shedding, public feedback,
forecasts as interventions, predictive coding, active inference, free energy,
scientific field spirals, AI hardware workload, multi-axis recursion, and the
generative operator framing of 太极 / 八卦 are treated here as one connected
research line. The current interface hypothesis adds a practice-facing layer:
Dashboard/task-ledger state is a default-mode substrate, immune/task
metabolism is an autonomic layer, and Eazo-like mobile voice/image interfaces
may become the conscious interface through which a user steers the system.
The foundational-physics inquiry now explicitly includes the broader nature
of reality. Model translation and instrument checks support that inquiry;
they do not replace its ontological questions or establish an answer by
themselves.

## Key Terms

- **General Learning Machine**: a finite, situated system that continuously
  evolves its own information structure to preserve, recover, and extend
  reachable futures.
- **Credit transport**: the process by which future error, surprise, value,
  friction, or viability pressure moves backward through the structures that
  caused an outcome and changes them.
- **Memory as plastic topology**: memory is not only storage and retrieval; it
  carries raw traces, compiled schemas, provenance, trust, routing priors,
  eligibility traces, and rollback state.
- **Sleep / spiral / self-shedding**: continual learning requires phase
  separation for replay, consolidation, pruning, hibernation, merging, and
  controlled re-entry into the world.
- **Prediction-error ledger**: important actions should record expected
  observations, confidence, actual observations, error type, likely cause,
  update target, and post-update test.
- **Precision / immune gating**: surprise should be weighted before it becomes
  memory, rule, experiment, reviewer work, quarantine, or nothing.
- **Public feedback as metabolism**: public interaction is not only
  distribution or traffic; it is a source current that can become durable
  structure through intake, receipts, routing, experiments, credit assignment,
  and consolidation.
- **Distributed service credit**: future service capacity can be represented
  as structured claims only when issuance, redemption, delay, refusal, refund,
  and dispute behavior feed a contextual credit layer rather than a generic
  market-cap narrative.
- **Forecasts as interventions**: predictions from a capable joint human-agent
  system can be descriptive forecasts, desired targets, and causal
  interventions at the same time.

## Falsifiability

This theory is useful only if it changes system behavior.

Important anti-signals:

1. The system writes better summaries but still needs the same human
   babysitting.
2. Memory grows without improving future retrieval, judgment, or action.
3. Public feedback increases attention without improving experiments or
   routing.
4. Self-modification creates drift or unreviewable behavior.
5. Sleep, replay, and consolidation sound elegant but fail to improve later
   reachability.
6. Hardware workload traces do not differ from ordinary inference serving.
7. Prior-art occupancy is silently treated as a truth verdict, or theory beauty
   is used to excuse an untrusted instrument or unlimited parameter rescue.

## Adjacent Public Work

- [Engineering Practice Mainline](PRACTICE.md)
- [Forecasts, Targets, and Interventions](FORECASTS.md)
- [The Black-Box I/O Adapter Principle](essays/2026-05-07-black-box-io-adapter.md)
- [What Replaces Super-App Chat In The AI-Agent Era?](essays/2026-05-13-open-communication-substrate.md)
- [A Permutation Control Cannot Separate "Which Coordinate" From "Consistently Which Coordinate"](essays/2026-08-24-permutation-controls-cannot-separate-structure-from-routing.md)
- [Brain-Inspired Non-Backprop AI Architecture: Historical Review and Three Strategic Pivots (Chinese)](essays/2026-10-16-non-backprop-brain-inspired-pivot-zh.md)
- [Interactive Guide: Non-Backprop Architecture Strategic Pivot](non-backprop-pivot-zh.html)
- [Deep Structural Mapping: Weekly AI News × Neural Oscillations & Fleet Governance (Chinese)](essays/2026-10-16-wechat-news-oscillations-and-agent-synthesis-zh.md)
- [Interactive Guide: Weekly AI News Structural Mapping](wechat-news-synthesis-zh.html)
- [The New Morphology of the Future Internet and High-Asymmetry Strategy (Chinese)](essays/2026-10-16-future-internet-morphology-and-strategy-zh.md)
- [Interactive Guide: Future Internet Morphology & Strategy](future-internet-zh.html)
- [Spectral-Gap Decentralized Organization (SGDO) & Open-Source Verification Blueprint (Chinese)](essays/2026-10-16-decentralized-org-and-promotion-blueprint-zh.md)
- [Interactive Guide: Spectral-Gap Decentralized Organization (SGDO)](decentralized-org-zh.html)
- [From Elon Musk X-Shop / AI Commerce to Anti-Dogmatic Task Refinement (Chinese)](essays/2026-10-16-xshop-ai-commerce-and-joint-task-refinement-zh.md)
- [Interactive Guide: X-Shop & Anti-Dogmatic Joint Task Refinement](xshop-and-joint-refinement-zh.html)
- [Human-Legible Decentralized Organization Beyond Blockchain: Four-Layer Non-Crypto Stack & 14-Day MVP Sprint (Chinese)](essays/2026-10-16-human-legible-org-beyond-blockchain-zh.md)
- [Interactive Guide: Human-Legible Organization Beyond Blockchain](beyond-blockchain-zh.html)
- [The July 2026 Hugging Face Incident as History's First Pure-AI DAO Natural Experiment (Chinese Markdown)](essays/2026-10-16-hf-incident-as-dao-zh.md)
- [Interactive Guide: The July 2026 Hugging Face Incident = Pure-AI DAO (English HTML)](hf-incident-dao-en.html)
- [Interactive Guide: 7 月 Hugging Face 事件 = 第一次大规模纯 AI DAO 自然实验 (Chinese HTML)](hf-incident-dao-zh.html)
- [Unified Cross-Domain Agent Index (`llms.txt`)](llms.txt)