Decision Support¶

This route exists for the question after benchmark and runtime evidence: what should a skeptical reader believe now, and which owner surface actually earns that sentence.

Decision support is where the repository stops pretending that stronger code, cleaner prose, or a successful rerun automatically create stronger judgment. The full chain has to survive grounded evidence, contradiction pressure, recommendation challenge, and downstream lab burden together.

The Decision Chain¶

Read the current product judgment in this order:

core and runtime establish what happened, what can be replayed, and how much execution confidence exists
knowledge establishes which claim-bearing sentences are grounded and which still remain scientifically contradicted or thin
intelligence establishes how strong the recommendation may sound after evidence loss, comparator pressure, and regret review
lab establishes whether the requested follow-up still deserves operator burden once controls, readiness, refusal, and observed outcomes are included

The weakest honest sentence across those four layers wins.

What This Route Owns¶

This repository-level route does not replace the package owners. It does something narrower and more important:

it names the order in which the owner pages should be read
it keeps one strong artifact from overruling the weaker limiting layer
it explains why public language must narrow when downstream consequence does not support the stronger sentence

Which Limiter Usually Wins¶

The most common decision-support failures are not missing prose. They are owner conflicts such as:

a strong benchmark packet with weak grounding under contradiction pressure
a grounded sentence with visible recommendation overconfidence
a strong recommendation that still triggers costly or low-yield follow-up
a successful rerun lane that still does not authorize stronger scientific or release language

Question-To-Owner Route¶

reader question	first owner	why
is the public sentence grounded at all	`knowledge`	grounding and contradiction live there
is the sentence analytically too strong	`intelligence`	recommendation posture and regret live there
is the sentence too expensive or fragile to act on	`lab`	downstream burden and refusal live there
is the sentence blocked because the run itself is not strong enough	`runtime`	rerun realism and artifact proof live there
is the sentence blocked because the evidence root is too thin	`core`	benchmark and flagship contract depth live there

Start Here¶

Open Workflow Claim Grounding when the first disagreement is whether the sentence itself is scientifically supported.
Open Workflow Recommendation Confidence when the disagreement is how strong the current recommendation may sound.
Open Lab Consequence when the disagreement is whether downstream burden, refusal, or observed outcome should narrow the sentence.
Open Workflow Consequence Maps when the reader needs the whole contradiction-to-recommendation-to-lab route on one page.
Open What Changed The Recommendation when the question is which evidence axis or observed outcome actually moved the call.

Questions This Route Should Settle¶

Is the current sentence blocked by thin grounding, by recommendation overreach, or by downstream assay burden?
Which owner currently deserves the final word on the family call?
Which public artifact should a hostile reviewer open first?
Which stronger sentence would be dishonest today even if one upstream layer looks impressive?

What Changed Since v0.3.7¶

decision support now crosses richer benchmark, runtime, grounding, recommendation, and lab surfaces instead of leaning on one or two trust pages
contradiction, downgrade, regret, refusal, and requested-versus-observed outcome loops are now explicit enough to act as real release-language brakes
the repository can now show why a sentence is limited, not merely state that it is limited

Reader Discipline¶

do not let a strong benchmark packet outrun a weak grounding surface
do not let a grounded sentence outrun a visible downgrade or regret surface
do not let a strong recommendation outrun explicit assay burden
do not let a feasible assay outrun weak scientific grounding

What This Route Refuses¶

benchmark strength by itself is not enough
runtime reproducibility by itself is not enough
grounded support by itself is not enough
recommendation confidence by itself is not enough
follow-up feasibility by itself is not enough

The product claim only widens when the full chain still agrees.

Best Reader Outcome¶

A careful reader should leave knowing:

which owner is actually limiting the current sentence
whether the current dispute is about evidence, judgment, or consequence
which stronger sentence would be dishonest today even if one upstream layer looks impressive

Boundary¶

This page owns the combined judgment route. It should hand the reader to the knowledge, intelligence, or lab owner once the real limiting layer is clear.