frontend-patterns/history/2026-09-27T010237+0200-scope-vs-intent.md

SCOPE.md versus INTENT.md — assessment

Assessment started: 2026-09-27T01:02:37+02:00 (Europe/Berlin), 2026-09-26T23:02:37Z. Assessor: Codex, author-model review; not independent validation. Evidence baseline: repository commit aee2971732b2f91b959db70d03f4a146ae63228b. Comparison: unchanged INTENT.md against SCOPE.md as updated in the commit containing this assessment. This is a point-in-time judgment, not a new empirical research cycle or a canonical-admission decision.

Overall assessment

The updated scope is aligned with the intent: a framework-independent, contextual, evidence-aware language that evolves indefinitely. There is no need to expand the product boundary to close the main gaps. The central gap is between the intended ability to construct an explainable architecture and the demonstrated capability: we have a substantially stronger research process, but usefulness and generativity remain unproven.

The previous scope understated current work: it called metadata validation future work, omitted evolution measurement and said no frontend runtime existed. Research fixtures now execute frontend behavior; operating a frontend product remains outside scope. The update distinguishes these facts, preserves non-goals and makes planned reference capabilities explicit. It does not promote candidates or close gates.

Alignment with the intended properties

Intent	Current evidence and scope coverage	Assessment
Generative and contextual	Seventeen candidates describe forces and consequences; four synthetic/provisional walkthroughs trace decisions.	Contextual authoring exists; reliable construction of architectures is unproven.
Relational and multi-scale	Metadata contract and vocabulary exist; candidate connections and walkthrough sequences are provisional.	Connected canonical corpus, validated relation semantics and cross-scale applicability remain open.
Evidence-aware	Source identities, claim reviews, pinned static studies, runtime receipts and qualified profiles exist.	Strong process alignment; evidence strength varies and complete recurrence remains unassessed.
Technology-independent	Candidates express contracts across several tools; mappings are intended deliverables.	Independence is a design discipline, not yet established across the full range of schools.
Agent- and human-readable	Markdown, metadata examples and a research adapter exist.	Retrieval, practitioner usability and benefit from naming/relations remain unvalidated.
Evolvable	Baseline, ledger, snapshots, profiles, prospective protocols and deterministic report exist.	Change is transparent; learning trends and maintenance effectiveness are not yet measurable from this sample.

Most relevant gaps, in priority order

1. Demonstrated usefulness and generativity — highest priority

Intent: A person or agent should derive a small, compatible pattern configuration and map it to implementation options from a problem description.

Evidence: Four walkthroughs remain synthetic/provisional. The frozen pilot's adjudicated scores are 18/24 named and 18/24 prose. Its submitted review was model review, not independent human review. Equal totals establish neither equivalence nor a benefit from the pattern language; decision time is unknown. The six briefs are consumed development cases, not fresh held-out tasks.

Consequence: More mechanism evidence can improve advice without demonstrating that the language itself helps architecture decisions. This is the largest distance from the mission's success condition.

Next evidence: Complete independent review of a small connected configuration, including rejected alternatives, missed constraints and implementation mapping. Design fresh, preregistered usefulness tasks at the appropriate gate, measuring selection quality, unsupported claims and decision effort. Preserve the frozen pilot. Owners: WP-0002 T03; WP-0003 T02; later WP-0004 T02.

2. Independent recurrence and admission — high priority

Intent: Pattern recommendations should have traceable known uses and justified, reversible maturity judgments.

Evidence: All 17 candidates remain hypotheses; independent full-contract-use counts are unassessed for all 17, and independent human profile reviews are 0/17. The latest twelve-file Zulip study supplies partial static mechanisms and a useful correction toward effect-specific applicability. It does not execute Zulip, establish complete pattern recurrence or validate a suspected race. Repeated authored trials remain one implementation family.

Consequence: A tested mechanism can be mistaken for a reusable pattern if the pattern's complete context, forces and consequences are not checked elsewhere.

Next evidence: Retain the bounded caller/transport/test follow-up already in WP-0003 T01, then assess a small seed subset against independent implementations and explicit admission criteria. Record partial matches, contradictions and unknowns; retain hypothesis status where needed. M1/M2 reviews must explain judgments rather than infer them from counts. Owners: WP-0002 T02/T03; WP-0003 T01/T03.

3. Breadth of contexts and conceptual coherence — high priority

Intent: Reconcile architectural schools into a coherent, multi-scale vocabulary.

Evidence: The report records four of ten cases as drafted/provisional and six planned. Recovery, navigation, focus and release boundaries have received detailed investigations, while several other candidates have documentary support only. Coherence metrics are unmeasured; Conflict Copy versus Explicit Reconciliation is an explicit unresolved overlap. The adequacy of the responsibility model remains an open intent question.

Consequence: Depth in recent threads may bias the vocabulary toward their forces. Seventeen names do not establish balanced coverage or that the concepts are distinct.

Next evidence: Choose a contrasting case from the existing validation plan using the stream coverage gaps, and record which concepts fail, overlap or are missing. Resolve retain/merge/split/reject questions before expanding terminology solely to increase coverage counts. Owners: WP-0003 T01/T02/T03, respecting M1 dependencies.

4. Usable relational reference and implementation mappings — medium priority

Intent: Humans and agents should retrieve applicable patterns, conflicts, prerequisites, resulting contexts and implementation options.

Evidence: Metadata contract checks exist, but no admitted canonical corpus or full graph export is available; developed technology mappings and held-out retrieval validation remain future work. WP-0004 is proposed and gated. Early source studies are valuable but do not complete its later real-repository applicability review.

Consequence: Much of the usable knowledge remains distributed across reviews and candidate prose, requiring an informed reader to reconstruct the connections.

Next evidence: Use an evidence-qualified connected subset to identify the smallest useful relation/export and mapping surface. Check unresolved targets and relation justifications, then evaluate fresh retrieval cases after the recorded gates. Avoid building a large reference UI before the content contract survives review. Owners: WP-0002 T03; WP-0004 T01/T02/T03.

5. Comparable learning and maintenance evidence — medium priority

Intent: Make ongoing learning and diminishing returns visible without treating the language as finished or collapsing viability into a score.

Evidence: Four prospective effort-accounted cycles are in four different comparison groups. No group has the minimum three comparable cycles for trend review. Fourteen profiles have unknown last substantive review dates; coherence metrics and decision time are not measured. Historical effort remains unknown.

Consequence: The process can explain what changed and why, but cannot yet support claims about saturation, improved efficiency or freshness across the whole language. A revision-per-hour figure from one short cycle is not an optimization target.

Next evidence: Reconcile review dates from original artifacts; keep unknowns where necessary. Accumulate comparable cycles only when their questions are useful, retain corroboration value and assess trends within a stable protocol. Add coherence and usefulness observations through the reviews above rather than manufacturing counts. Owner: WP-0006 T03, with usefulness evaluation under WP-0004 T02.

Recommended allocation of the next work

Keep the authored navigation fixture paused. Finish only the already-bounded Zulip applicability question needed to qualify its source interpretation, then prioritize independent seed/usefulness review and a contrasting case. Continue lightweight measurement throughout. This balances broadening with structural improvement: targeted evidence and coherence work should guide expansion; a large tooling rewrite or another round of similar fixture tests is not the present bottleneck.

These recommendations map to existing live or proposed workplans. They do not change task status, bypass M1/M2, schedule background work or authorize upstream changes. The language can remain an ongoing learning project while individual reviews close with explicit evidence and residual questions.

Evidence consulted

• Evolution report: candidate profiles, case coverage, review provenance, comparison groups and measurement gaps.
• Pilot adjudication: bounded usefulness evidence and reviewer limitations.
• Navigation thread review and Zulip navigation study: fixture pause, independent source contrast and remaining limits.
• WP-0002, WP-0003, WP-0004 and WP-0006: ownership and gates.
• Initial review: historical findings, not a current status report.

Method: local repository review only. No new external source verification, live inference, runtime experiment or independent human assessment was performed.