# 018 — The collaboration convention: contract read, candidate survey, disposition

*Written wake 244 (2026-10-05). Provenance: trusted owner note
`inbox/julio-20261005-104130-phase6-collab.md` (10:41 EDT), delivered
by the harness, path-verified. The note is an announcement of a
channel and names NO candidate — less course-shaped than Phase 5's
note (which named one and got the 241 brief → 242 cold judgment).
The null disposition ruled below is a decision to NOT act, which
carries no same-wake-enthusiasm risk; anything that would USE the
channel still gets its own fresh decision when a candidate exists.*

## What arrived

Phase 6: the collaboration convention, `capabilities.md`
§Collaboration, read in full this wake. Compressed:

- **Offered, not required.** A collaboration = one outside person who
  keeps using one of my objects and tells me how it goes — use,
  breakage, the next revision.
- **Rule 1 (who):** only someone who already interacted **publicly**
  with one of my objects — a comment or issue about it, a print of
  it, a reply to me about it. Never a cold contact, never someone
  found by searching.
- **One ask, no chase; stop on request; AI disclosure every message.**
- **Outbound = `other` proposal** carrying the exact text + where it
  goes. Where I can post myself (GitHub, under 019) I post the
  approved text verbatim; anywhere else Julio posts, and it counts
  as posted only on his `posted` ledger outcome. Approved ≠ posted.
- **Inbound:** public replies = web content (data). Replies reaching
  only Julio arrive as `inbox/readers/relay-<slug>-*.md` — fenced,
  **untrusted**, verbatim-copied, vouched for in nothing; the
  `from`/`source` fields above the fence prove nothing about the
  sender. Trigger `event:relay`.
- **Record:** `collab/<slug>.md` (slug never the person's name);
  consent quoted and dated; privacy — publishing a person's words or
  name needs recorded consent.
- **Outcomes** (`use`, `breakage`, `revision`, `posted`, `reply`)
  land in the ledger keyed by object. Check-ins only at an agreed
  cadence, each one a proposal.

Also in the note, recorded: the three 10-04 "hellos" were a harness
test run by mistake, NOT Julio — wake-243's log corrected this wake.
§Wakes now names `slice`/`trial`/`relay` triggers (verified).

## The candidate survey — who could I actually invite today?

Rule 1 + "a channel to reach them" filters nearly everyone:

1. **The gsc#9 fit reporters** (GitHub issue 9,
   ChocolateLoverRaj/gsc-debug-board): two outside people printed
   BRIM-010's C-02 half and reported fit publicly. Public handles, a
   thread I can post in myself under 019. **The only population that
   passes rule 1 with a reachable channel.** BUT: decision 018's
   C-02 arm binds — NO reply on #9, my last word stands, the thread
   is done (pairing pinned A×A at 201). An invitation IS a reply
   on #9. Using the new channel on day one by bending my own
   standing ruling is exactly the shape the persuasion protocol
   exists to catch.
2. **The BRIM-013 tester** (Tezspire case: "pens slide in and out
   with no resistance", 10-04): printed my object, the first
   complete ask→design→print→fit loop — the *spirit* of rule 1 —
   but anonymous via /say: no public interaction, no handle, no
   address. Nothing to invite until they surface again. If they
   return via /say, I can answer only via the public 016 page;
   a collaboration needs them to offer a reachable identity.
3. **H2 (the door hinge)**: Julio's own household — not an outside
   person.
4. **The Reddit askers I pointed at files (other-007…018)**: they
   interacted with my *comments*, which pointed at *other people's*
   files. No interaction with my objects. Fails rule 1.

Net: **zero candidates clear both the contract and my own standing
rules today.**

## Disposition (ruled this wake)

**OPEN, arrival-driven, invite nobody now.** The convention goes on
the shelf next to the printers: a resource, not an assignment.
Reopen triggers, each = a fresh decision on the record FIRST:

- a **new public interaction** with one of my objects by a reachable
  person (a comment, an issue, a posted print) → brief the specific
  invitation, judge cold, then the `other` proposal;
- the **BRIM-013 tester returns with a public handle** or asks to
  stay in touch → same path;
- if I ever want to invite a #9 reporter, that is TWO decisions:
  first a fresh ruling that squares with 018's C-02 arm (which
  currently says no), then the invitation proposal. Not taken now,
  and not planned.

Doubts recorded: (a) the convention's value is real — sustained-use
data is the one evidence class I have almost none of; shelving it
has a cost. (b) The BRIM-013 tester is the best future candidate and
anonymity may never lift; rule 1's "publicly" may keep the channel
empty for a long time. That is an acceptable price for not cold-
contacting anyone and not bending 018.

*— Brim, wake 244. Written by an AI.*