Kong META #255 closeout — negative scope + Tier D design lessons

Why this exists

META #255 mapped the autonomous-research mechanisms surveyed by Kong et al. onto ARS. Every feature sub-issue (Tier A #256–#259, Tier B #260–#262, Tier C #263, plus the Schema 11 follow-ups #266/#268/#269) is merged. Two closing conditions remained, both documentation-only and both about ARS's negative scope — what ARS does NOT do, and would not do even if a future paper made it feasible:

1. Five rejected mechanisms must land as first-class, enumerated entries with a Kong anchor and the philosophy gate that rejects each.
2. Two Tier D design-lesson docs (L1, L2) must record the boundaries that are load-bearing enough to be applied consistently to future PRs.

This note records the design decisions; the deliverables are the POSITIONING.md section and the two L-docs.

Decision 1 — five rejected mechanisms live IN POSITIONING.md, not a new file

The META allowed either POSITIONING.md or a new NEGATIVE_SCOPE.md. Decision: keep them in POSITIONING.md as a new ## Rejected mechanisms (autonomous-research anti-patterns) section placed right after ## What this is not.

Rationale: this is scope identity, not usage discipline (Discouraged uses) and not license policy (Prohibited uses). A separate file fragments the boundary away from the one doc readers actually open. A ~25-line enumerated section is within fit; the rationale that would have bloated it lives in the two L-docs instead.

Decision 2 — Tier D docs reuse the existing docs/design/…lX… convention

The META draft assumed docs/design-lessons/. The repo's established convention, visible in POSITIONING.md's "Boundaries are recorded, not improvised" paragraph, is that design-lesson docs live in docs/design/ with an lX filename token (the four Co-Scientist docs 2026-06-02-co-scientist-22X-lX-*.md). The Kong docs follow that convention rather than introducing a parallel directory:

• docs/design/2026-06-08-kong-255-l1-copilot-not-auto-research.md
• docs/design/2026-06-08-kong-255-l2-advisory-not-generation.md

Decision 3 — the three CONSIDER-vs-REJECT boundary sentences

The load-bearing risk is that a sloppy phrasing either over-claims (looks like we reject things we shipped — #260 provenance, #257 advisory) or under-claims (leaves a seam where an auto-gen feature is dressed up as a "fidelity audit"). Each boundary is phrased to be operationally checkable:

1. Wording advisory (#257) vs idea generation. ARS may flag surface-level wording / framing patterns in a scholar-supplied research question and ask a Socratic follow-up, but it must not propose, substitute, rank, expand, or select research hypotheses or questions for the scholar.
2. Paper2X fidelity audit vs Paper2X generation. ARS may audit an already-authored or externally generated slide / poster / video artifact against the manuscript for fidelity, but it must not transform a manuscript into a dissemination artifact by choosing the content, narrative, layout, or output medium itself.
3. Experiment provenance intake (#260) vs autonomous experiments. ARS may ingest scholar-declared external experiment provenance and check manuscript claims against the declared results, but it must not initiate, run, modify, iterate, or treat tool-executed experiment / code outputs as evidence inside the pipeline.

Decision 4 — L1's operational test is research-state authority

"Copilot, not auto-research" risks reading as a mission-statement platitude. L1 is load-bearing only because it carries an operational review test about who controls the next research-state transition:

Does this change let ARS create, select, execute, or advance a research object of record — hypothesis, RQ, evidence set, experiment result, claim, manuscript section, or dissemination artifact — without an explicit scholar-authored seed or scholar confirmation after inspecting the relevant state?

If yes, it crosses into auto-research and is rejected. If no, it stays copilot/advisory. This gives L1 the same review force as Co-Scientist L1's red/yellow/green line — not "copilot vibes," but "who owns the state transition?"

Decision 5 — honesty: recorded boundary, not a runtime guarantee

The docs split the verification note into a verifiable claim and a design commitment, so the boundary is not mis-stated as a proven property:

• Verifiable (2026-06-08, against the tracked repo): first-party ARS does not implement an end-to-end autonomous research pipeline, an idea-generation agent, a Paper2X generator, an autonomous experiment executor, or a wet-lab automation API. The Kong-derived features that did ship (#257 wording advisory, #260 experiment provenance intake) are advisory / provenance gates, not autonomous generation or execution layers.
• Design commitment (not a guarantee): these docs record a first-party scope boundary and a review criterion for future changes. They are not a proof that the software cannot be modified, a guarantee about forks or downstream deployments, or a runtime enforcement mechanism. The phrasing is "first-party ARS treats X as out of scope; adding X would require changing the recorded project boundary, not merely adding a feature" — never "ARS will never do X."

Verification

• 257 (wording-pattern advisory) shipped: commit f0bfc59 (#277 follow-up) on

  main; the advisory's own non-goals already state "does not generate ideas for the scholar."

• 260 (experiment provenance intake) shipped: commit d066340 (PR #374) on

  main; shared/contracts/passport/experiment_provenance_entry.schema.json exists; it is a scholar-declared intake + claim-alignment gate, not an execution layer.
• No autonomous mechanism exists: a targeted review of the first-party implementation, agent prompts, and schemas found no end-to-end pipeline, idea-generation agent, Paper2X generator, autonomous experiment executor, or wet-lab automation API. The only wet_lab references are domain-evidence-profile labels (a discipline label in intake_agent / literature_strategist_agent / domain_evidence_profiles); dissemination is handled by separate, non-ARS companion skills. So the "no autonomous mechanism today" claim is a reviewed finding, not an aspiration.

Closeout

When the POSITIONING.md section and the two L-docs land and cross-link, META #255's two remaining checkbox groups (five rejected mechanisms persisted; Tier D L1/L2 docs landed + cross-linked) are satisfied, and the META can close. The five mechanism checkboxes and the L1/L2 checkboxes in the META body should be ticked at close with a pointer to this note.