DUNIN7 · LOOMWORKS · RECORD
record.dunin7.com
Status Current
Path standing-notes/loomworks-personal-memory-posture-note-v0_1.md

Loomworks — Personal-Memory Posture Note — v0.1

Version. 0.1 Date. 2026-08-16 Status. Working draft. Records an Operator decision of 2026-08-16. Consumer. The Operator; future sessions touching the Companion, recall prioritization, the catalog, or the capabilities document; Claude Code for the verification item in §5. Provenance. Operator direction, 2026-08-16: with personal memory being claimed by the platform players (Apple and peers), Loomworks should be unencumbered from the responsibility; and the capability requires analysis and AI token work, making it a costly sideline without real benefit. Decision confirmed against Claude's scalpel-not-axe lean: retire the ambition, retain the substrate. Prior positions engaged. The personal-memory and proactive-Companion investigation (Phase 41–45 arc framing); the catalog v0.1 entries 38–43; the capabilities document v0.1 "personal recall" framing; the critique v0.1 §2.4/§4 non-viability finding. Preserved alongside, per discipline. Canonical home. loomworks-record. HTML primary, this Markdown as source.


Plain-language summary

Personal memory is retired as a Loomworks product ambition and retained as substrate. The platform players own the consumer memory space, the critique already ruled it non-viable, and the Operator adds the deciding cost argument: personal memory consumes analysis and AI token work with no commercial return — a costly sideline. What stays is the person-scoped engagement as working context: the professional's preferences, who's-who, and delegation machinery that make the Companion useful inside real engagements, spending tokens only when explicitly invoked. The recall build effort re-aims accordingly: engagement recall first, personal inheriting whatever it builds. One verification item goes to Claude Code: inventory anything on the personal path that spends tokens without explicit invocation, and surface it for retirement.


1. The decision

Retired: personal memory as product ambition. Loomworks does not market, demonstrate, or invest polish in remembering a person's life. Apple, OpenAI, and Google claiming that space is treated as a gift: it commoditizes a layer Loomworks should not fight for and sharpens positioning by contrast. The external line: your assistant remembers you; Loomworks remembers the work.

Retained: the person-scoped engagement as substrate. One scope in the N-space model, carrying the professional's working context — preferences, who's-who, delegation grants, cross-engagement continuity for the Operator. Architecturally cheap to keep, expensive to extract, and the proving ground where the truthfulness disciplines (server-composed replies, human-only destructive gates) were established before protecting every engagement.

2. The three grounds

Market. The critique v0.1 already ruled personal memory non-viable against free assistant-native memory; the platform players' arrival confirms rather than reopens it.

Cost — the Operator's deciding ground. Personal memory requires analysis and AI token work per use and invites background enrichment ambitions; with no commercial return, that spend is pure sideline. The recent defect record carries the same lesson from the effort side: a large share of the truthfulness arc (the B-81 class, B-107, change requests 199 through 220) was spent polishing the personal capture path — professional-grade correctness applied to a consumer-grade purpose.

Risk. Personal-life data is the erasure-rights-heavy class where the critique's P3 tension (memory that never forgets versus law that requires forgetting) bites hardest. Shrinking the personal ambition shrinks P3 surface before the retention-and-erasure investigation even begins.

3. What the substrate retention means operationally

The retained scope is passive-cheap by rule: it spends tokens only on explicit invocation — a capture the person makes, a recall the person asks. No background analysis, enrichment, summarization, or proactive processing runs against personal Memory. Spend that does occur runs through the credit gate like every pipeline operation. The ambition's retirement is what makes the retention cheap; the two halves of the decision fund each other.

4. Consequences

  1. Catalog v0.2 rider. Entries 42–43 (endurance training, the puppy) drop to demonstration status. Entries 38 and 41 (chronic condition, estate readiness) survive reframed as what they always were — succession cases whose consumers are other people at a harder time, the business thesis in home clothes. Entry 40 (family history) sits between; Operator's call at catalog v0.2.
  2. Capabilities document, next bump. "Personal recall" reframes from life-memory to working context — a professional capability, still honestly Live. The maintenance rule's next triggered review carries it.
  3. Recall prioritization sharpens. With personal deprioritized, the recall effort's first target is engagement recall — where the record's strongest lead already points (the never-scoped semantic-recall foundation effort was carved out of the engagement-side path), and where every vertical's buyer actually lives. The personal side inherits whatever engagement recall builds; it drives nothing. This note is the source for the build-list re-aim; the queue ruling itself ("recall next," and now specifically engagement-recall next) remains the Operator's pending word to Claude Code.
  4. Positioning cleanly inverts a competitor into a neighbor. Per the curated-ingestion investigation's cooperation logic: assistant memory could one day be a source a person commissions ingestion from — a friendlier relationship with the platform players than competition, recorded as an echo, not a plan.

5. Verification item for Claude Code

Inventory the personal path for token consumers that run without explicit invocation: proactive behaviors (the Phase 44 machinery), classifiers, enrichment, or scheduled work touching person-scoped Memory. Report what exists, what each costs per occurrence, and which are candidates for retirement or gating under this posture. Read-only inventory; retirement itself is a subsequent Operator-gated change request if anything is found.

6. What this note does not do

It does not remove the person-scoped engagement, the personal capture pathway, or personal recall — all remain Live substrate under §3's passive-cheap rule. It does not amend the seed; the personal scope is architecture-specification-level and this is posture. It does not close the truthfulness disciplines proven on the personal path; those are permanent and scope-general. It does not decide catalog entry 40. It does not pre-decide the build-list ordering; it supplies the ground for the Operator's ruling.

7. Drafter's note

v0.1 produced by Claude under Operator direction, 2026-08-16, same session as the critique and ingestion documents it cross-references. The two-sense distinction (ambition versus substrate) was Claude's lean; the cost ground and the confirmation are the Operator's. Prior positions revised — the catalog's personal entries as products, the capabilities document's life-memory framing of personal recall — are named in §4 with their revision paths; the Phase 41–45 personal-memory investigation stands as the trajectory that built the substrate this note retains.


DUNIN7 — Done In Seven LLC — Miami, Florida Loomworks — Personal-Memory Posture Note — v0.1 — 2026-08-16