The Panel API Contract
How This Book Works
Continuum Press is a living technical publication. Every interactive panel on this site follows one published contract. This page is that contract, in plain language.
The Panel API Contract
Continuum Press is a living technical publication. Every interactive panel on this site follows one published contract. This page is that contract, in plain language.
Evidence
Reader
Continuum Press is a living technical publication. Every interactive panel on this site follows one published contract. This page is that contract, in plain language.
Two roads, and they never merge silently:
Evidence: Netlist → Run → Trace → Verification → Publication Reader: Run → Exploration → Comparison → No evidence promotion
A claim becomes ESTABLISHED only by traveling the first road, in the verification pipeline, with the results stored and versioned. A reader's run can never jump roads — not automatically, not ever. If a reader's experiment deserves to be in the book, it is re-run, verified, and published with its own manifest.
| Tag | Meaning | Color |
|---|---|---|
| ESTABLISHED | Reproducible, run-manifest-backed, verified by the pipeline | blue |
| READER_EXPERIMENT | A live run you made in the browser. Exploration, not evidence | grey |
| SIMULATED_ONLY | A simulated topology with no physical board behind it (tagged, never hidden) | blue, distinct |
Every panel carries a chain-of-custody block: model version, SHA-256 hashes of the netlist and the trace, verification date, verifier, and the frozen solver regime. The hashes are computed from the exact bytes of the downloadable artifacts. If your downloaded netlist hashes differently, the artifact you have is not the artifact we published — write to us and we will fix it publicly.
Every simulation in every chapter can be downloaded and re-run offline in ngspice (free). Each evidence bundle includes a REPRODUCE.txt script and expected values for self-checking. Where a chapter's numbers were ever wrong, the erratum is published on the page, not buried.
Every panel states what its model is and what it does not claim. The Chapter 3 PDN panel, for example, is a lumped RLC network — trustworthy for anti-resonance behavior up to a few hundred MHz; results outside that range are exploratory only. A model that cannot say what it cannot do is not evidence.
Live panels expose parameters through .param value substitution only. You can change component values inside published ranges; you cannot change topology, solver settings, or the published baselines. The fixed parts are the experiment's controls — the same discipline Chapter 5 applies to physical boards.
Your runs happen entirely in your browser. They are logged locally (your browser's storage only) and are never sent to our servers, never stored, never counted. The Reader Trust Index in the footer says "reader experiments: not tracked" because it is true.
A panel is allowed on this site only if it passes: a model version ID; a frozen, documented solver regime; parameters via .param substitution with clamped ranges; defaults that reproduce the ESTABLISHED traces; grey READER_EXPERIMENT overlays on blue baselines; an insight card separating computed (from your sliders) and observed (from your trace); a visible model-limits note; downloads in sync with what the panel runs; and, where physical claims are near, an honest status object.
This contract is itself versioned with the platform. Last update: 2026-10-02.