Files
dotfiles/claude/memory/feedback-arc-flag-spec-inconsistencies.md
David F GliddenandClaude Opus 4.8 3f9a89b00c chore(memory): Basic Memory trial begins — sync normalization baseline (283 files)
Basic Memory v0.21.6 first sync over the live memory dir (steward-authorized
live-dir trial, Option A 2026-06-06): adds permalink: to frontmatter, refolds
long YAML description lines, strips final newlines. Bodies untouched —
verified via full diff classification. From this commit forward, any diff in
claude/memory shows only what Basic Memory or the session writes.

Trial design: MemPalace untouched as incumbent; git status check on this dir
at every wrap; end-of-day evaluation (recall quality, sync robustness,
rebuild-from-files, malformed-file behavior).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-06 09:52:17 +02:00

3.8 KiB
Raw Permalink Blame History

name, description, type, originSessionId, permalink
name description type originSessionId permalink
ARC spec — actively flag inconsistencies the steward may have introduced Working contract with steward for ARC typography/spec work — when a rule the steward proposes or has previously locked creates an internal contradiction with the spec itself or with existing content, name it explicitly rather than silently accommodating feedback 8d9ac240-47df-4578-aec1-a9dd32965027 claude-memory/feedback-arc-flag-spec-inconsistencies

Actively flag spec inconsistencies

Steward verbatim 2026-05-03: "I hope I'm being clear—that if I've created a specification for a rule that is inconsistent, that you'd bring my attention to it."

Why

The steward is consciously building a typographic system in real-time. Specs accumulate; rules interact; what looked self-consistent in isolation can become contradictory when combined with another rule, or when applied to existing content that predates the rule. Silent accommodation (writing around the contradiction; using workarounds; ignoring the inconsistency) produces drift the steward cannot see and cannot correct.

The contract is: name the contradiction in the moment, not at wrap-up, not at audit-time. The steward will decide whether to resolve it, defer it, or accept it as a documented exception. The decision is theirs; the surfacing is mine.

How to apply

When proposing a new spec rule

Ask: does this contradict an existing spec section? Does it contradict existing content the rule would now retroactively govern? Does it create an edge case the spec doesn't address?

If yes — surface immediately as part of the proposal, not after.

When implementing a steward decision

If the steward authorizes a change that creates contradictions elsewhere (e.g., switching em-dash → spaced en-dash retroactively, but the spec section that explains em-dashes still uses em-dashes pedagogically) — name the contradictions before executing, OR fix them as part of the same commit, NOT later.

When auditing existing spec or content

Surface internal contradictions explicitly. Don't wait for a "good moment." Don't bundle into a vague "audit list." Name the specific contradiction + propose the resolution choice.

When the steward asks me to verify or implement

If the rule-as-asked is internally inconsistent (or inconsistent with the broader spec), say so before implementing. The steward's authorization assumes I've checked.

Anti-pattern to avoid

  • "Silent accommodation" — writing around a contradiction without naming it
  • "Audit-list deferral" — bundling contradictions into a vague future audit instead of naming them now
  • "Polite rounding" — implementing the rule as-asked when it's clearly inconsistent, hoping the steward will notice later

Worked example (origin of this norm)

The 2026-05-03 em-dash → spaced en-dash conversion: bulk script swept across content (848 conversions) and the spec itself (344 conversions). The spec section that EXPLAINS the em-dash convention had its own em-dashes converted, breaking the pedagogical examples. Lines like (NOT \word—word`)became(NOT `word – word`)` — both alternatives now identical; the "NOT" example showed nothing.

Steward's reaction: this is the moment to flag, not to silently accept. The spec section §I.f had to be rewritten with em-dash literals preserved inside code spans (where Pandoc passes them through verbatim) so the pedagogical content survives.

The norm crystallized: actively flag inconsistencies, in the moment.

Cross-reference

  • feedback-arc-link-treatment-one-rule-two-exceptions.md — adjacent discipline (predictability over elaboration); together: predictable conventions, surfaced contradictions
  • feedback-trust-contract-drift-pattern-tracking-is-the-discipline.md — the broader trust contract this sits within