This publication brings together the working protocols of the Semantic Cascade Method developed in 2026. The complete texts can be read here on the site; the original files are also available to download.
Start with the two current operating branches below: the complete 4D specification and the shorter 0.4B protocol. The version history records earlier trials and revisions, keeping the instructions for a full run separate from the work actually carried out.
This library preserves the 2026 development record of the Cascade-Semantic Method. It contains two current operating branches, nineteen earlier version documents, and one lineage guide. Historical trials, copied controls and editorial revisions are not presented as completed independent experiments.
Current operating branches
- 4D — Cascade-Semantic Method: Complete Operating Specification — current long branch, 80,352 planned for current frozen run
- 0.4B — Portable English Protocol for the Cascade-Semantic Method — current short branch, Q=8K
Version history and lineage
- Version lineage and limits — Editorial lineage guide; not a separate method execution.
- Reference pipeline — expanded 72-question local cell — Executed local reference slice: 72 questions and 30-sentence answers, 2,160 answer sentences. Not a full all-object/all-frame run.
- 0.3D — short Russian D answer profile — Executed editorial short-form slice; first hard character enforcement was superseded by later soft-target/no-rewrite policy.
- Transitional English 4D pilot — translated short D material — translation/editorial pilot, 24 local questions
Earlier local experiments and revisions
- A/B A — object-and-triad expansion (historical 2025 mechanism) — Historical protocol; 7,830 planned question positions for the revised 15-object run, 564 answers recorded. Not a completed full A run.
- A/B B — initial 36-question whole-prompt pass — Executed historical pilot; its asserted relation among three influencers was rejected by the operator and must not be treated as task fact.
- A/B B — corrected 36-question whole-prompt pass — Executed: 36 questions and 36 five-sentence answers. Distinct from the early invalid B pass.
- B36 — whole-prompt refocus on receiving income — Executed local 36-cell refocus; separate from 0-series and later profit wording.
- 0 — six operations in factual and counterfactual modes — Executed local 12-question pilot; later labeled the group-A reference for zero-series tests.
- 0A — unshifted-prompt control — Executed control: 12 questions; 10 of 12 answers were copied literally from the reference and two edited. Not an independent repeat.
- 0B — creation-focused prompt — Executed: 12 newly worded answers of five sentences on the same local cell.
- 0C — triple question expansion with literal first-child kernel — Executed local expanded run: 36 questions, 15 sentences per answer.
- 0D — compound six-interrogative question beginnings — Executed local expanded run: 36 new compound questions and freshly formulated 15-sentence answers.
- 0.2 — return_004-derived document baseline — Executed local document-derived slice: 8 questions, 10 sentences each.
- 0.2A — copied unshifted-prompt control — Executed control by copying all eight baseline questions and answers, not by independent model generation.
- 0.2B — creation-focused short document branch — Executed local slice: eight fresh ten-sentence answers; later editorial corrections are a separate revision.
- 0.2C — three children and thirty-sentence answer expansion — Executed local slice: 24 questions and thirty sentences per answer.
- 0.2D — thirty-sentence compound-interrogative document branch — Executed local slice: 24 questions and 30-sentence fresh answers; superseded editorially by revised 0.2D.
- 0.2B revised — operational acceptance and continuity correction — Executed editorial revision of eight-question material; not an independently regenerated mechanism test.
- 0.2D revised — operational criteria and reduced rubric repetition — Executed editorial revision of 24-question material; not a new blinded trial.
These are research and operating documents, not the books or the June 2025 White Paper. The source Markdown is unchanged in the downloadable archive. A limited test cell is not a completed full method run.
Download the original documents
Download the complete original ZIP (22 Markdown documents and 2 JSON metadata files; SHA-256: 902c773425579207f2e62730e0ca8f54285f65525784c80ded50775d677c4511).