Skip to main content

Equation study — the hypothesis stamp

The framework doesn't only run known equations — it can generate them: the seven-step wizard synthesizes an equation from a query, operators can be derived, and the community can propose new ones. That capability comes with an obvious risk to integrity: a generated equation could be mistaken for a discovery. The framework refuses to let that happen. Every equation it generates is intrinsically labelled a hypothesis — stated by the code itself, before any interpretation, so nobody has to think to ask "but is this real?"

The hypothesis stamp​

The stamp is free, code-level, and involves no language model. It is attached to a generated equation as a plain fact about where the equation came from:

{
  "label": "hypothesis",
  "not_a_discovery": true,
  "statement": "Framework-generated research hypothesis for study — NOT proven physics, not a measurement, not a discovery. No result is asserted; verify independently.",
  "subject_kind": "wizard",
  "generated_by": "compute master equation",
  "provenance": { "…": "what produced it" },
  "zeqond": 2296532592
}

That statement is the single source of truth for the wording — it says the same thing everywhere a generated equation appears. The subject_kind records what produced it:

wizard

An equation synthesized by the seven-step wizard from a query.

derived-operator

An operator derived from the existing catalogue and constants.

community-operator

An operator proposed through the community process, not yet canonical.

Because the stamp is provenance plus a fixed sentence — not a judgement — it's deterministic and identical on every node, and it's stamped with the Zeqond it was made at like everything else.

Opt-in study, still honest​

Sometimes you do want deeper analysis of a generated equation. The second layer is an opt-in study that runs only when a human explicitly asks for it, over your own language-model key (BYOK). It is advisory and nothing more:

  • it is hard-framed as studying a hypothesis — the model is instructed to say "this hypothesis appears to…", to flag uncertainty, and never to declare a result true;
  • it never asserts truth and never mutates the canonical operator registry — a study can't promote a hypothesis into accepted physics;
  • it's yours: it runs on your key, on your request, and the framework's own stance on the equation stays exactly what the stamp says.
# create a study of a generated equation (opt-in, over your own LLM key)
curl -sX POST https://zeqsdk.com/api/equation-study \
  -H "Authorization: Bearer $ZEQ_KEY" -H "Content-Type: application/json" \
  -d '{"subject":"…","equation":"…"}'
# read it back
curl -s https://zeqsdk.com/api/equation-study/STUDY_ID

Why this belongs in the core​

This is the honesty contract applied to generation. The same framework that refuses to fabricate a compute result refuses to let a generated equation pose as an established one. New ideas are welcome — the framework will happily synthesize them — but they arrive wearing an honest label, and turning a hypothesis into accepted science is a road the framework will not walk for you.