Short answers about the agent and family packs, each quoting a pack's own documents or pointing to them.
No. In the words of each pack's license: “This package is a normative specification. It is not software, SaaS, a managed service, consulting, support, managed operations, implementation delivery, or a validation service.” The specification defines how a conforming agent, or family, must behave; your engineering team builds the agent, or the family, from it, in any language or paradigm. The only executables in the package are the reproducibility binaries of the NNT (Non-Normative Tests) companion — observational evidence, not normative, and not a reference build.
Because the specification is the product. A reference build in one language would become the thing to copy, and the specification would stop being execution-neutral. Each license is explicit: the package “is not a reference implementation, and no reference build is shipped.” The five reconstructions in the NNT companion are study artifacts — evidence that the specification can be built from the pack alone — not implementations to follow. The pack asks you to read them only after your own implementation passes the Conformance Tests.
Python, Rust, Prolog, C and Bash are five different ways of building software: a high-level scripting language, a systems language, a logic language, an imperative language and a shell. For each agent, and for the SSCAR family, five paradigm-isolated reconstruction contexts built it from its pack alone, one in each, and ran the pack's conformance records against each reconstruction; the outcome for each pack is reported on its agent page or on the family page, and in its NNT companion. The five contexts were isolated from one another but shared their reconstruction model, prompt authorship and training data, as each companion discloses. This is observational evidence that the specification is derivable and self-sufficient — not a certification, and not a list of supported stacks. The supplementary derivations are described in the next question.
Each agent pack, and the family pack, has a separate supplement that carries the same specification into five further target paradigms: Common Lisp, Brainfuck, spreadsheet formulas, an Analytical Engine-style program and AWK. They were chosen to span very different eras and models of computation: the Analytical Engine design was described in 1843, Lisp dates from 1958 and Brainfuck from 1993, while the comparator rendering is in Rust, announced in 2010. The renderings are built from the same published Specification and Conformance Tests; depending on the paradigm, they run a bounded sweep of the agent's or the family's state transitions, compared byte for byte with a reference output, or the full public conformance suite, or both. The supplement is not part of the specification and not part of the NNT companion: it has its own Manifest, which names the pack Manifest as its parent, so it can be verified on its own. It is observational evidence that the specification carries over into very different models of computation — not a certification. The Analytical Engine itself was never built; its program runs on an emulator.
Check it first against the pack's own instruments. Each license describes its agent, or the family and its components, as semantically closed: “every term is resolved within the package's Semantic Index and Ambiguity Register.” The Semantic Index lists the pack's normative terms, the Ambiguity Register resolves its lexical overloads, the Conformance Tests carry records anchored to the specification, and the Verification Runbook in the package (verify.md) has you reproduce the pack's declared values on your own machine: a mismatch is a fact you report, not a judgment you make. If the error stands, write to contact@semantichor.com.
None are promised. The license states that “SEMANTICHOR is not obligated to provide support”, and that the package is provided “as is.” It covers one version — “the package version identified by the manifest_ref” shown on each agent's page and on the family page — and “confers no right to updated, future, or other versions.” A revised specification is published under its own license.
Within its license, yes. The package is distributed free of charge, and the license grants, “Subject to the Restriction section”, among other rights, “the right to deploy and operate those implementations for any lawful purpose, including in the course of a commercial business”. It also reserves uses, in its Restriction section. Read both in full: each license is published on this site, and each agent's page, and the family page, set out what they permit and what they reserve.
An agent pack. The author constructed the pack's seven canonical files, including the conformance tests, reviewed them manually, and designed and calibrated the NNT harness. Large language models were used openly, as tools to accelerate research and production, with manual review: they assisted the drafting of the pack's documents and, by design, built the five reconstruction baselines in the NNT companion, which are study artifacts, not reference implementations. No tool is an author; the author is accountable for the content. What matters is not which tools touched the documents but whether the pack can be verified: its integrity is anchored by the Manifest and its conformance suite is runnable.
The family pack. The author constructed the canonical files of the six units of the family package, including the conformance tests and the connector annexes, reviewed them manually, and designed and calibrated the NNT harness. Large language models were used openly, as tools to accelerate research and production, with manual review: they assisted the drafting of the package's documents and, by design, built the five reconstruction baselines in the NNT companion, which are study artifacts, not reference implementations. No tool is an author; the author is accountable for the content. What matters is not which tools touched the documents but whether the package can be verified: its integrity is anchored by the Manifest, and its conformance suite and connector tests are runnable.
The supplementary derivations. The author designed and calibrated the harnesses of the derivations in each supplement. Large language models were used openly, as tools to accelerate research and production, with manual review: they produced the renderings (Common Lisp, Brainfuck, spreadsheet formulas, an Analytical Engine-style program and AWK) and the comparators (the Rust rendering and the reference output) in an AI-provider environment, as study artifacts; they are not reference implementations. No tool is an author; the author is accountable for the content. What matters is not which tools touched the files but whether they can be verified: each supplement has its own Manifest, which names the pack Manifest as its parent.