Physics from the records observers can share.
Physics preprint · Pragma Research · Français
Observer Patch Holography (OPH) investigates whether familiar physics can be reconstructed from bounded observers that reach agreement. An observer patch has local state, a boundary with ports, readback, records, and feedback for repairing disagreement. Within the model, facts become public when records survive comparison across overlapping patches.
The proposed universe is one structure viewed both as a timeless consistent history and as the internal local computation of that history. Its records and responses to interventions supply the operation; no external host or preferred global clock is required. Space, physical duration and matter are intended emergent readouts. Current constructions supply some geometry, states, actions and clock maps as premises. The interpretation and working terms distinguish this physical proposal from the conditional results achieved in these constructions.
This repository contains the scientific work behind that idea: papers, machine-checked proofs and executable models. It connects finite agreement with quantum probability, spatial geometry and the symmetry pattern of the known forces, with a conditional route to Einstein dynamics. Each construction states the assumptions it consumes. Establishing a common physical realization requires measurements that distinguish the proposed models.
We have conditional mathematical results and concrete finite repair candidates. The candidates reproduce and connect some required structures, and we can distinguish several of them in simulations. We do not have a common model established as physically correct. The candidate register and model search separate constructed results, supplied assumptions and missing physical connections.
OPH's physical correctness is a hypothesis under investigation. Logical consistency, conditional theorems and successful simulations do not establish that nature follows OPH or any particular candidate repair law. We retain alternative laws and revise or reject physical attachments when tests fail.
The finite constructions supply model comparisons. A physical test additionally needs an operational identification of states, records, time and parameters; a distinguishing observable; an account of calibration and uncertainty; and a prediction and rejection rule fixed before independent comparison data are examined. Compare against established physics and alternative models, including cases where they cannot be distinguished. The physical-test programme and frozen-prediction ladder track that work. Reproducing a value used to choose the model is a fit, not an independent test.
The parallel programme-wide obstruction search seeks contradictions, joint-realization impossibility, unavoidable independent physical inputs and empirical exclusions. Each result states which fixed formulation or model class it rejects. A decisive negative result is valuable; failing to find one does not confirm OPH. The issue also distinguishes scientific refutation from a practical decision to stop pursuing the programme.
Find the simplest shared repair rule that satisfies the known constraints.
The programme is complete only when that shared repair rule has been found and shown to satisfy the required constraints on one common model. Component proofs, simulations and compiler improvements are milestones toward that result; none closes the programme by itself. Physical identification also requires discriminating evidence.
The research priority is an explicit microscopic repair law whose evolution supplies the inputs of the existing theorems on one common model. Separate conditional results can be correct without their physical premises being realized together. We start with the simplest compatible construction, reuse proved results, and test each added hypothesis against alternatives. The carrier geometry, reversible gauge response and local operational clocks must be compatible readouts of that same process. Add a mechanism only for a demonstrated missing connection, and count every added physical assumption. A rejected scalar-only candidate does not establish that a simple common source law is impossible.
Use the established agreement mathematics as the research foundation: specify recovery and scheduling assumptions, map linear overlap models to sheaf diffusion and boundary extension, and reuse the resulting theorems where their premises hold. Each candidate must identify its remaining physical obligation and a test that can distinguish its law.
Compare the concrete candidate laws, their tested properties and remaining choices. Follow the model search, its model–theorem premise map, and the research issues. Progress means a tested connection or a rejected candidate; theorem counts do not measure physical completion.
Paper releases and broad prose restructuring are on hold until a coherent model milestone supports them. Audits, correctness fixes and reproducibility checks continue. The research policy states the milestone and the correction exception.
Main-branch PDFs are current-source previews and may retain the existing release identifier after corrections. Tagged or deposited release artifacts retain their original bytes; a preview is not a new release.
| To explore… | Start with… |
|---|---|
| The model being constructed and its open physical connections | Model search |
| The active repair laws, assumptions and distinguishing tests | Candidate register |
| The existing physics argument | Finite Observer Consensus as a Reconstruction Principle |
| The proofs and reproducible calculations | Lean library and reproduction guide |
| The related work on learning machines | Cadence and interactive demos |
| Explanations for a general audience | Pragma Research blog |
The research program connects these parts of physics through one observer architecture, with separate assumptions stated for each result.
- Agreement between observers. Under the stated termination and consistency conditions, local repairs reach a protected public record independent of their order. The finite theorems describe agreement, stability and refinement.
- Quantum probability. Finite event algebras support the Born probability rule, Lüders conditioning and exact Tsirelson results, including a Bell-state construction that attains 2√2. These are mathematical results on the stated event-algebra branch.
- Geometry and the known forces. A twelve-port response construction supplies a three-dimensional spatial readback. Complete reversible response and internal observer transport force the Standard Model gauge Lie type under the stated hypotheses. A supplied matter representation gives an anomaly-free fifteen-state generation; its physical realization and the global gauge group require additional structure.
- Spacetime and gravity. Declared source and reading laws support controlled causal-geometry constructions. The Einstein-equation implication assumes a common physical realization with the stated stress, entropy, continuum and scale data.
- Classical and quantum matter. A supplied charged-scalar/Maxwell action supports controlled nonlinear continuum motion in its real sector and an interacting quantum state space on a fixed mesh. Selecting the action from observer histories and identifying it with physical matter are separate requirements.
- Constants as fixed-point problems. Koide's relation holds exactly under a stated balance premise. The fine-structure calculation certifies a root of a declared closure map and carries diagnostic status. The capacity program asks whether the public capacity assigned to the universe agrees with the capacity reconstructed from within it. Connecting these constructions to measured constants requires physical identification.
The main preprint develops these connections and their precise scope. The paper index provides the specialist accounts, including thermodynamics, field geometry and fixed-point constructions for the constants.
The Lean library contains more than 12300 public theorems and lemmas with no admitted proofs. Exact certificates and reproducible simulations accompany the mathematical arguments. The companion physics simulator provides executable observer dynamics and retained evidence.
The axiom reference states the three core axioms; the premise register and model–theorem premise map record additional premises and result-specific assumptions, including physical identifications and empirical inputs. The postdiction ledger records comparisons with measured values and their input ancestry; the frozen-prediction ladder records tests whose conditions must be fixed before the comparison data are examined. The falsification program gives the observations that would refute particular claims.
After setting up the dependencies in REPRODUCE.md, check the claim graph and selected finite algebra, record-capacity and consensus results:
python3 tools/check_claim_registry.py
python3 -m pytest -q \
code/a5_closure/test_audit.py \
code/capacity_readback/test_correctable_public_record_capacity.py \
code/capacity_readback/test_reversible_public_checkpoint_packet.py \
code/consensus/test_reference_architecture_benchmark_suite.py \
code/consensus/test_verified_tree_packet_net.pyThe reproduction guide describes the wider checks and the scope of each evidence family.
The OPH reconstruction map connects observer records, three-dimensional source geometry, causal order, clocks, fields and quantum states. Each arrow names a mathematical connection; the papers state the assumptions that join them into an effective physical description.
| Path | Contents |
|---|---|
flagship/ |
Main standalone physics paper, TeX source and PDF |
paper/ |
Core papers and publication index |
Lean/ |
Machine-checked mathematical development |
code/ and evidence/ |
Executable models, certificates and reproduction evidence |
extra/ and cosmology/ |
Focused mathematical and physical research papers |
book/ |
Reverse Engineering Reality, source and downloadable book |
docs/ |
Canonical references: axioms, frozen predictions, postdictions and falsifiers |
docs/registers/ |
Generated premise, observation, instrument and selection registers |
docs/research/ |
Markdown research notes, each verified by a code lane |
docs/policies/ |
Data, hardware-evidence and writing policies |
Prefer a simpler shared rule over an accumulation of separate mechanisms.
Contributions should repair an existing defect or resolve a named uncertainty
in the common model and its tests. Reuse proved results; add a proof when a
specific model obligation needs it. Standalone finite-theorem expansion and
theorem-count targets are outside this programme. Start with the
research issues
and reproduction guide. Keep exploratory logs local and report
concise decisions with their assumptions and evidence. Verified research notes
belong in docs/research/; paper publication follows the hold above.
Cadence makes the settling idea executable as a learning architecture. Its bounded software patches carry local state and readback, with feedback that repairs prediction errors. Its default brain, System 1, learns from experience as animal brains do and not by backpropagation. Optional System 2 observers add recursive feedback inside the same settlement.
The Cadence preprint describes the architecture and experiments. Its computational case rests on learning behavior and measured resource use. Pragma Research connects this work to embodied AI.
OPH has an official memecoin on Base, TEO (Theory Of Everything). Its token
address is 0xd2a7055317D0c7b316319cDadaE592D0644A0b07
(chart on Dexscreener).
Any other token using the OPH or TEO name is unrelated to this project.
The token was set up so that Bernhard Mueller can claim all the trading fees it generates, and those fees have been a very helpful source of funding for OPH. The token has no role in the OPH research.
Nothing here is financial advice.
The repository uses split licensing. All software, including the Lean library, code/, tools/, and the schemas in schemas/, is licensed under Apache-2.0. Papers, the book, documentation, figures, data, the generated ledgers in tracking/, and the packaged particle data in pdg_data/ are licensed under CC BY-NC-SA 4.0; the tabulated values in pdg_data/ carry their upstream Particle Data Group terms. The LICENSE file gives the per-directory map.