14 branch-closure rows and 12 reviewer-v2 rows close the current finite state without a physical claim.
ITHKOR-SIM
Spend compute where information is forming
ITHKOR-SIM is a new practical research direction. It does not treat ITHKOR as a finished physical theory; it treats it as a principle for more efficient simulation.
The goal is to allocate compute budget by coherence, response, records and information pressure, not only by fixed grids or hand-tuned thresholds.
What it brings
- Many simulations spend effort uniformly: every pixel, cell, particle or step follows the same schedule.
- ITHKOR-SIM asks whether more compute should go where stable structure or strong response appears.
- Checkpoints should form where durable records appear and future steps or observers will need them.
- Compression should happen only where negative controls show that important behavior is not being erased.
QPU rows remain explicit-approval-only. The website should not sell them as physics proof.
At the same 25% update budget, ITHKOR-SIM beat five non-oracle controls across eight deterministic trials.
Temporal-jet checkpoint records passed in counted structured toy worlds; the noise-drive row remains a separate uncounted stress row.
Observable-capsule compression beat source, gradient, delta, uniform, random and shuffled capsule controls in counted worlds.
The result allows a real/semi-real simulation pilot, not a claim of better physics.
Practical pivot
From branch closure to adaptive simulation
D23 does not add a physical claim. It makes a more useful move: ask whether finite diagnostics help inside a practical simulator.
D23 names supported islands, blocked bridges and negative controls.
Instead of forcing another physics bridge, it asks whether diagnostics improve compute allocation.
SIM1 compared policies at the same 25% update budget.
Uniform, gradient, source-only, random and shuffled controls had to stay visible.
SIM2C/SIM3C changed the object: temporal-jet records and observable capsules instead of bare checkpoints and masks.
The point
Not simulating the universe, but making better choices inside a simulator
ITHKOR-SIM is not a claim that the world is a simulation. It asks a more practical question: if compute budget is limited, can we spend it more intelligently than uniform updates or simple geometric heuristics?
The existing ITHKOR corpus provides finite diagnostics: coherence, source-response, record stability, internal event progress and approximation risk. The SIM direction tests whether those diagnostics can become useful simulation policy.
SIM1
The first toy result is small, but practical
SIM1 ran in a synthetic 2D scalar toy world. Every policy could update only 25% of cells per step. The ITHKOR-SIM policy combined model-delta pressure, source-response, record stability and coherence pressure.
Result: ITHKOR-SIM beat uniform, geometric-gradient, source-only, random and shuffled-information controls across 8 deterministic trials. The best non-oracle control was source_only, and the margin over it was 0.074887.
That is a first small signal of practical value. It says nothing about physical reality. It only says that information-adaptive compute allocation can preserve observables better than simple controls in this toy setup.
SIM2C/SIM3C
Checkpointing and compression now have a narrower practical pass layer
The original SIM2/SIM2B and SIM3/SIM3B results are not erased: bare checkpoint placement and mask-only compression remain boundaries. That is why SIM2C/SIM3C matter: they change the engineering object, not the claim.
SIM2C adds a temporal-jet checkpoint record: a checkpoint stores state plus a local time derivative, and reconstruction uses Hermite interpolation. SIM3C adds an observable capsule: a small mean/peak metadata layer above a hybrid detail mask.
Both layers passed 3/3 counted structured toy worlds. Stress/noise rows remain separate and uncounted. Publicly, this means we have a candidate for a real/semi-real simulation pilot, not a universal simulator and not physics proof.
Why D23 helps
Negative branches become safety brakes
D23 is not only a positive list. It preserves MOD1 as a negative anti-pass-hunting row and repeats that motion, equivalence, energy, gravity, spacetime and General ITHKOR claims remain blocked.
For SIM, that is useful. An adaptive policy should not only chase attractive scores; it should also know when it is outside scope. A good practical simulator needs boundaries as much as signals.
SIM results
What is a pass layer and what remains a boundary
The SIM branch is most useful when positive layers and negative boundaries stay visible together.
| Layer | Result | Careful reading |
|---|---|---|
| SIM1 adaptive update | 5/5 non-oracle controls | At 25% update budget, the compute/error tradeoff improved in a synthetic finite toy world. |
| SIM2/SIM2B bare checkpoints | 0/3 + 0/3 | Simple event-only or coverage-plus-event checkpoints did not beat the uniform-time baseline. |
| SIM2C temporal-jet records | 3/3 counted | The checkpoint stores state plus local derivative; this is a candidate policy for a pilot, not a physical claim. |
| SIM3/SIM3B compression | 0/3 + 2/3 | Mask-only and hybrid detail rules exposed boundaries that must not be renamed into a broad pass. |
| SIM3C observable capsules | 3/3 counted | A small metadata capsule layer helped preserve observables in counted structured toy worlds. |
| Stress/noise rows | separate | Noise-drive rows are not counted as passes and should remain visible in the next pilot. |
Claim boundaries
SIM is an engineering branch, not physics proof
The page is more credible when the practical direction stays separate from large physical claims.
We test whether finite diagnostics help allocate compute, checkpoints and compression in simulations.
SIM1 is a first small engineering signal at fixed update budget with visible controls.
SIM2C/SIM3C allow pilot testing of temporal-jet checkpoints and observable-capsule compression with controls.
ITHKOR-SIM does not claim that the universe is a simulation or that ITHKOR is a finished physical theory.
It does not replace physics engines and does not confirm gravity, spacetime, physical time or quantum gravity.
D23 claim boundary
What D23 and SIM1 do and do not imply
The strongest public sentence is practical: finite diagnostics may be useful as adaptive policy. There is no need to inflate it.
| Topic | Status | Public wording |
|---|---|---|
| D23 reviewer v2 | Ready | Documents the current finite state and the practical SIM pivot. |
| QPU rows | Locked | QPU repeats are explicit-approval-only and not an automatic website claim layer. |
| MOD1 | Negative row | Remains a negative anti-pass-hunting witness. |
| SIM1 | Toy pass | ITHKOR-SIM improved the compute/error tradeoff in a synthetic finite toy world. |
| SIM2C | Scoped pass | Temporal-jet checkpoint records passed 3/3 counted structured toy worlds. |
| SIM3C | Scoped pass | Observable-capsule compression passed 3/3 counted structured toy worlds. |
| Physics | Do not claim | No proof of gravity, spacetime, physical time or quantum gravity. |
Technical bundle
Download the technical and reviewer bundle
The public page explains ITHKOR in readable language. The technical bundle contains the more precise documents: experiment atlas, claim ledger, reviewer brief, the D20-D23 replication path and the ITHKOR-SIM layer including the SIM2C/SIM3C practical pass.
ithkor-public-reviewer-bundle.zip
ZIP, 41 KB, public-safe Markdown documents
Download full ZIP bundle
SHA-256 2506B65A142E7ADA932DFCC4D364CC91452D3DF4C187C4405F40D0B8233D4494
Start simply
Begin with the big question without heavy mathematics or inflated physical claims.
Technical check
For readers who want to inspect terminology, branches, tables and claim boundaries.
Technical reader, Markdown
Master Theory v6
The main ITHKOR v6 synthesis as a finite information-response program.
Open document Reviewer, Markdown Scientific VersionA drier technical version with methodology, controls, boundary taxonomy and limitations.
Open document Atlas, Markdown Experiment AtlasBranch overview: question, result, boundary, control, QPU status and claim frame.
Open document Claim audit, Markdown Claim LedgerThe safety document separating allowed finite claims from blocked claims.
Open documentReviewer path
The shortest route for skeptical inspection: what to open first, where the replication gates live and how D23 closes the current branches.
Reviewer, Markdown
Reviewer Brief
A short entry point for skeptical readers: what to inspect first and how to read QPU results.
Open document D20, Markdown D20 Reviewer BundleCross-branch package of representative rows: positive islands and the explicit negative MOD1 result.
Open document D21, Markdown D21 Reviewer SmokeManifest, raw-folder, QPU-lock, claim-audit and MOD1-stays-fail checks.
Open document D22, Markdown D22 Drift ReportFresh-clone rerun: status reproduction of non-QPU rows, QPU explicit-approval-only and separate hash drift.
Open document D23, Markdown D23 Reviewer Bundle v2Current branch-closure and reviewer-v2 state: 14 closure rows, 12 reviewer rows, QPU explicit-approval-only and no physical claim.
Open document Ledger, Markdown Branch Closure LedgerA map of supported finite islands, blocked bridges and negative controls before the ITHKOR-SIM pivot.
Open documentPractical SIM path
A practical branch: use finite diagnostics for adaptive simulation, checkpointing and compression without turning it into a physics proof claim.
SIM, Markdown
ITHKOR-SIM Research Program
A research program for information-adaptive compute allocation, checkpointing, compression and negative controls.
Open document SIM1, Markdown SIM1 Adaptive Budget ToyFirst toy test: at the same 25% update budget, ITHKOR-SIM beat 5/5 non-oracle controls across 8 deterministic trials.
Open document SIM2C/SIM3C, Markdown SIM2C/SIM3C Practical PassPractical pass layer: temporal-jet checkpoint records and observable-capsule compression passed 3/3 counted structured toy worlds.
Open document
Implications
Why this is a good product direction
Next steps from ITHKOR-SIM
SIM is the practical branch. Continue with the experiment map or download the reviewer bundle and MCP packages.