chore: final public-repo hygiene — remove internal process, counsel work product, and commercial strategy - #166

Replaces 58 first-name attributions with "the maintainer" in 11 documents. The
decisions, dates, and their force are unchanged; only the identifier is.

Three legal drafts additionally dropped the line stating that the maintainer "is
the reviewing counsel." Those documents already carry, and keep, the accurate
statement that they are drafts pending outside counsel review; asserting who is
standing in as counsel in the meantime is a disclosure the drafts do not need to
make and that a reader could reasonably use against the project.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

The four legal documents were published as drafts that announced their own
unreviewed status: DRAFT in the titles, status banners naming the review that had
not happened, launch gates written as counsel deadlines, and a "Notes to counsel"
section at the foot of each. A reader looking for Helpthread's licensing terms
found a work-in-progress instead of an answer.

They now read as what they are: the project's terms. Removed from all four are
the DRAFT titles, the status and gate banners, and the counsel-notes sections.
Every open legal question those sections held is preserved, in more detail than
the documents carried, in HT-104 through HT-109.

The Module API Exception keeps one operative line, because its status is a fact
about the grant rather than a disclosure about process: an AGPL-3.0 §7 additional
permission takes effect when it is appended to LICENSE, and it has not been
appended. Publishing it as though it were in force would grant permissions the
project has not decided to grant, and a granted permission cannot be withdrawn
from copies already conveyed. legal/README.md gains a matching "What is and is
not in force" section in place of its counsel-deadline list.

Nothing operative changed in any of the four. The grant text, the license terms,
the trademark rules, and the provenance commitments are byte-for-byte what they
were.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

…xcerpted

An independent adversarial review rated the previous arrangement a blocker: the
"not yet in force" banner sat above the exception, and the exception text below
it is written as a complete, standalone license addendum. Anyone quoting the
"exception text" section — which is the natural thing to quote — carried the
grant without the qualifier.

The failure that follows is concrete. A module author points at the published
text, ships a proprietary combined work in reliance on it, and is out of AGPL
compliance because the permission was never appended to LICENSE. The project
then inherits a dispute that starts with "your own repository said we could."

The notice now sits inside the quoted block, ahead of the grant, and is marked
for deletion on adoption. The grant text itself is unchanged.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

…tory

The license governed nothing here. It applies to paid modules, which live in
their own separate repositories, and its own text says it replaces the
placeholder LICENSE in each of those. A proprietary subscription contract sitting
in the open-source tree describing artifacts that are not in the tree is noise at
best, and at this stage it reads as commercial apparatus arriving before the
product it governs.

It moves to ResonantIQ/helpthread-internal and, on adoption, into each paid
module's own repository. legal/README.md now says only that separately licensed
modules carry their own terms elsewhere, which is the fact a reader of this
repository needs.

The AGPL-3.0 §7 module API exception stays. That one is the opposite case: it
modifies this repository's own license grant, so it has to ship with the work —
a downstream who receives the core without it receives plain AGPL only. It
remains unadopted and says so inside its own quoted text.

What this repository now states about licensing is true of this repository: the
AGPL-3.0 governs everything in it, and one additional permission is drafted but
not in force.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

…them

The policy committed the project to specific techniques, in absolute terms,
forever. Techniques age. A published commitment to a technique becomes a stick to
beat the project with the moment a better one exists, or the moment a single
instance departs from it — and the commitment bought nothing that a commitment to
the outcome would not have bought.

Each method commitment is replaced by the property it was there to produce:

- "copyleft codebases are not opened during development" — the outcome is that no
  copyleft-derived code enters the shipping tree, which the policy already states.
  How that is achieved is an engineering decision.
- "license-verify every third-party dependency at adoption" — the outcome is that
  every dependency in the tree carries a compatible license, and that one whose
  license does not resolve does not ship. The timing of the check is not the point.
- "every substantive change receives real human design and review before it is
  merged" — the outcome is that nothing reaches main without review and the
  maintainer's approval, with the record preserved. How review is performed is
  explicitly not committed to here.

Two legal theories are also removed. The policy asserted that human review is what
supports the copyrightability of AI-assisted work under U.S. Copyright Office
guidance. That inference — review implies authorship — is contestable, it is not
the project's to assert, and the AGPL-enforcement position does not depend on it:
copyright claims belong to whoever owns the lines at issue, which section 1 still
says. The corresponding item in section 7's list goes with it.

The document opens by stating the distinction, so the next reader knows which
sentences are commitments and which are description.

CLAUDE.md carried a harder form of the same claim and is corrected to match.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>