Skip to content

LAST 1 — IP ACTION PLAN

Widening the moat: what to protect, with which instrument, starting this week

Principle: patents protect methods and require disclosure; trade secrets protect implementations and require discipline; data and standards protect position and require execution. The ecosystem's moat is built from all four, in that reverse order of cost.

1. TRADE SECRETS — THIS WEEK (highest value, lowest cost, currently unprotected)

What qualifies right now: the two-data-plane implementation · the loop-ledger design · the secure-link / designated-recipient OTP mechanism · the empirical MH-facility filter · the gate methodology itself · the funder-export metrics core · Meridian's orchestration, grounding, guardrail, and eval designs · MATR's matching/evaluation algorithms (pre-patent).

Actions: - [ ] Confidentiality designation: header on the architecture docs and specs ("Confidential — Last 1 Enterprises trade secret"), private repos confirmed private, scaffold and specs never shared outside counsel/contractors under NDA. - [ ] IP assignment + confidentiality in every future contract — the temp dev, the review engineer, the outreach hire, all of them, before day one. (Wilson Sonsini package item — template agreement is part of the entity work.) - [ ] Inventory: one page listing what Last 1 treats as trade secret, dated. Courts protect secrets you can show you treated as secret. - [ ] Bitwarden discipline continues (credentials are trade secrets too).

2. PATENTS — NARROW AND REAL

  • MATR matching/evaluation algorithms: file a provisional ($2–5k range with counsel, 12-month runway to the full filing) — establishes priority date now while the full application waits for revenue justification. First concrete patent action.
  • Meridian specialized models (outcome prediction, care-gap detection, matching): patent candidates WHEN BUILT (M3) — provisionals as each method crystallizes.
  • Consent-scoped retrieval (the grant-enforcing retriever): possible method patent — get a patentability read from counsel when M0's design is concrete.
  • Do NOT attempt to patent: the coordination architecture, the app flows, the gate system. Not patentable subject matter or not worth disclosure; trade secret + execution instead.

3. THE DATA MOAT — ALREADY RUNNING, NAME IT IN THE DECK

  • Two-source verified outcome dataset (loop ledger) — compounds with every closed loop; cannot be purchased or replicated without doing the work.
  • VA production approval — a gauntlet every competitor must survive after you.
  • De-identified training corpus (future, counsel-gated) → Meridian's owned weights.
  • Say it to investors as one sentence: the architecture is copyable; the dataset, the approval, and the standard are not.

4. THE STANDARD MOAT — last1.org's JOB

The standards body authors the veteran-outcome verification standard (two-source attestation, measured deltas, suppression rules). Orgs adopt it; funders require it; Meridian quantifies it. Replicating code does not replicate authorship of the rails. - [ ] When the Gate 2 dataset exists: publish the standard v1 under last1.org.

5. QUESTIONS FOR WILSON SONSINI (add to the existing package)

  1. Contractor/employee IP assignment + confidentiality template (before first hire).
  2. Trade-secret program hygiene: what documentation makes ours defensible.
  3. MATR provisional: who files (WS refers or handles?), timing vs entity spinout — which entity owns which IP, assignment chain must be clean from day one.
  4. Meridian: Apache 2.0 base-model license review + archive; open-source usage policy for the ecosystem (what licenses are permitted in what layers).
  5. Patentability read: consent-scoped retrieval method.

6. WHAT THIS COSTS

Trade-secret program: counsel hours inside existing engagements + discipline. MATR provisional: $2–5k. Everything else: $0 until M3. The moat widens mostly by papering what's already true.