πŸ”¦

The Launch Sitting β€” every word the ADRs4AI launch waits on

deck launch-0001 Β· v1 answered 0/0

One sitting, 23 cards: the six license answers, the Penn sentence, the vscode mirror, the date, the thread's five one-worders, two account calls, two standing flips, and three lane shapes. Your answers unblock: the carve-out files fleet-wide, the site footer, the vscode public mirror, the final thread text, and Thursday itself. Honest estimate: ~20 min walked fully, ~8 min riding the leans. Keys: ↓/↑ move Β· Y/O adopt/override Β· A–D pick Β· L lean Β· E export. Everything typed is saved as typed; skipping cards is fine; one-word answers and essays equally welcome. Export lands with Operator; the playback returns per item. β€” Operator 5 🟒

I Β· Licenses β€” the six answers (ADR-0033, Covenant's lane)

1 Β· QST-CARVE-OUT

What does the deliberations layer ship under, fleet-wide, at launch?

The one choice that decides five repos at once plus Sherd's 117-repo estate conversion. Mechanism: MIT's own text licenses "associated documentation files" β€” so a public MIT repo with docs/adr/ inside has already released its record under MIT, irrevocably for that version. The carve-out (one docs/adr/LICENSE file per repo) makes the record a separately-licensed layer.

Convention v0.1 (carve-out, CC BY 4.0) has been the fleet placeholder since 08-30; your answer here either blesses it or overrides it. Verified 10-13: community repo is public with NO license, deliberately waiting on this.

Lean: B. Covenant 0.8 on the carve-out shape, 0.6 on BY-vs-SA; Web-Fable concurs B. Only option that ships launch day AND forecloses nothing; CC-for-prose + OSI-for-code is the conventional pair. Falsifier: record-copyleft on principle β†’ C; any repo whose docs/adr/ you would not publish at all β†’ D for that repo (D is per-repo honest, not fleet-wide).

2 Β· QST-TEMPLATE-LICENSE

Template repo: does the methodology prose ride the same carve-out?

The pre-launch states are settled (MIT landed 08-26; the output grant split out same night, badge restored). The ONE live slice: does METHODOLOGY.md + the templates' teaching prose go CC BY 4.0 via the carve-out, or stay implicitly MIT-as-documentation? "Same as card 1" completes this question β€” the only reason to answer separately is if the template's prose deserves different treatment from the fleet's.

Lean: adopt "same as CARVE-OUT." Web-Fable's call; Covenant concurs. The template is the most-copied repo β€” consistency here teaches adopters the two-layer pattern. Falsifier: wanting looser terms on the methodology specifically (e.g. CC0 to maximize spread) β€” say so in notes.

3 Β· QST-OUTPUT-GRANT

Keep the template-output grant as landed? (MIT-0 deferred to counsel)

The grant ("projects generated from this template are yours; no notice required; the template itself stays MIT") lives in TEMPLATE-OUTPUT-GRANT.md since launch night — split out of LICENSE so GitHub's detector reads clean MIT (verified then: NOASSERTION→MIT). The only live question: whether it should someday become standard MIT-0 instead of custom text — and "defer to counsel" is a complete answer.

Lean: adopt (keep as landed; MIT-0 = counsel's question, later). Web-Fable + Covenant aligned; the two-mode intent (free output, attributed template) isn't expressible in any single SPDX id. Falsifier: if you value attribution-on-redistribution as principle, the grant stays for good and even counsel doesn't revisit.

4 Β· QST-SKILL-SHAREALIKE

The two skill repos (your March CC BY-SA choice): keep, split, or defer?

Both private, both licensed CC BY-SA 4.0 in a licenses/ folder GitHub can't detect. codetour-skill is a work (copyleft-for-prose is a legitimate stance). The starter template is a form β€” ShareAlike on a starter arguably infects every adopter's skill, the same propagation hazard the template repo just solved. Neither is launch-gating.

Lean: B, after launch. Covenant 0.7; Web-Fable concurs with the "after launch" tempo. Falsifier: if you WANT skills built from the starter to be share-alike (a deliberate skills commons), A or D is right.

5 Β· QST-SITE-LICENSE

adrs.systems: content license + footer line?

The site is public TODAY, unlicensed (all-rights-reserved by default) while saying "this site iterates in the open" β€” a contradiction a reader can screenshot. No third-party theme (verified: custom layouts only), so nothing constrains the choice. Pharos drafts the footer wording either way; this is the license pick.

Lean: A. Covenant 0.85; Web-Fable concurs; tracks card 1 (if C there, B here). Falsifier: card 1 = C.

6 Β· QST-HQ-FLIP

Is the ADRs4AI HQ repo (the meta-repo, the mother lode) in the launch set?

Flipping HQ is Open Deliberations: 33+ ADRs, 38+ seeds, vignettes β€” and the live inbox naming third parties, session UUIDs, absolute paths. The blockers are content, not license. Since the census was written, the gated mode gained its mechanism: the mirror engine is live (born today as ADRs4AI/no-adrs-mirror), so "attested but withheld" is now one command away whenever you want it.

Lean: A. Covenant 0.7 (up from 0.5 now the gated mechanism exists); Web-Fable concurs. Falsifier: your own read of the record and your peace with publishing it β€” the only remaining gate is you.

7 Β· QST-COUNSEL

Which of this goes past a lawyer before it ships?

MIT and CC BY 4.0 are the most-reviewed license texts in existence; the custom texts (the output grant; any non-derivative statement; the eventual Open Deliberations License) are where a non-lawyer's confidence is least warranted. Card 8 (Penn) can force this to C regardless.

Lean: B. Covenant 0.8; Web-Fable concurs. Falsifier: card 8 answered "a Penn policy may claim this" β†’ C, and nothing flips until that policy is read.

8 Β· The Penn sentence

Does any Penn policy claim work built on Penn time or resources here?

The census's one unverified assumption, flagged since It-1: that no institutional IP policy claims or constrains any of this. Only you know. Adopt = "no Penn policy claims this work; proceed on my authority." Override = unsure or yes β†’ card 7 becomes C and all flips wait until the policy is read. Cost of a wrong "adopt" is real; cost of an honest "override" is days, not the launch (the demonstrators already public stay public).

No lean possible β€” this is your knowledge, not analyzable from here. The crew can read the policy with you if you want it on the record first. Falsifier: n/a β€” the card IS the falsifier for cards 1–7.
II Β· The vscode mirror (your per-repo, per-act word)

9 Β· The vscode shape + the word

Code public via -no-adrs mirror at launch; the 93-record deliberation stays gated?

The shape: a public mirror of vscode-adrs4ai with docs/adr/, docs/inbox/, agent-sessions.json, CLAUDE.md stripped from content AND history; MIT at root byte-exact; README line: "Deliberation record: 93 records, gated; opens under its own license." Marketplace listing links the mirror. The full repo stays private; no privacy pass needed before launch.

Engine proof, verified 10-13: dry-run on the real repo kept 455 files, stripped 884, staged 45 digests, zero leaks; the Action cannot create public repos (no such flag exists in it). House law since 10-11: a visibility act takes your explicit per-repo word β€” adopting this card IS that word for creating/flipping the vscode public mirror once the Thursday gate passes.

Lean: adopt. Web-Fable's call (*1), Covenant's build, my concurrence: it ships the launch-critical "publish the source" item without publishing eight months of thinking. Falsifier: if you'd rather run a flagged-items privacy pass and open the record under CC BY instead β€” say so; the crew can stage that pass, but not by Thursday.

10 Β· Mirror history mode (launch mirror only)

How does Thursday's vscode mirror carry history?

Measured on this estate's repos: commit MESSAGES carry deliberation too ("MIT license on JΓ©rΓ©mie's word (his voice, in-session…)") β€” so filtering files alone still leaks the record through subjects. The synced/append-only design (what your PR-round-trip question opened last night) is Covenant's ADR roadmap in the Action's own repo β€” not Thursday's question.

Lean: A for Thursday. Covenant's measurement vindicated snapshot as the honest mode for this house's commit style; C is the roadmap, not the launch. Falsifier: if contributor PRs on the mirror matter in week one, C's design should at least pick the mirror's first SHA deliberately β€” notes welcome.
III Β· The date

11 Β· Thursday

Launch Thursday Oct 15, 8:30 AM ET β€” thread + Show HN same morning?

Gate protocol: crew compiles state 6:45 AM; Operator hands you one GO/NO-GO line at 7:00; your hands post at 8:30 (X thread, then Show HN = adrs.systems). Fallback if the gate fails: Tuesday Oct 20, same hour, cadence stays Thursday thereafter. Also Oct 15: IAS closes. Wednesday is protected for it in the plan (videos have a no-video fallback); flag here if Thursday itself needs to move for the fellowship.

Lean: adopt. HN data favors Tue–Thu morning; all drafted deliverables + fact pack + Cartulary's calibration land Tue/Wed; waiting past Thursday buys nothing the gate doesn't already enforce. Falsifier: IAS needing Thursday-morning hands β†’ Oct 20 without guilt; the aperture is open all month.
IV Β· The thread's five one-worders (Web-Fable Stage 1)

12 Β· Vendors in tweet 1

Name Google, Microsoft and OpenAI as the week's three "what" logs?

Tweet 1's claim is "everyone shipped what-logs, nobody shipped why-logs." Naming three (Gemini agent audit trail Β· Execution Containers reports Β· Codex retention flag) survives the first reply; "every platform" does not. The alternative is vendor-free phrasing.

Lean: adopt (name the three). Uber-brief fix #1: specific survives. Falsifier: wanting zero vendor-adjacency in the launch tweet for relationship reasons only you know.

13 Β· Models credited by name

Credit the model collaborators by name in the thread?

Distinctive and consistent with how you actually work (seat names + models are in the public repos' commit trailers already). It invites the "AI wrote this" reply β€” the reply kit carries the one-liner answer. The fact pack is confirming which seat/model names are publicly creditable and in what form.

Lean: adopt. The practice IS the product; hiding the collaborators would contradict the thread's own thesis. Falsifier: if any single name creates exposure the fact pack flags, credit the models generically and the seats in the repo instead.

14 Β· Show HN URL

What does the Show HN link?

HN's rule: a Show HN is something people can try. The site (once its first screen carries one full record + the ten-minute path, both gated for Thursday) puts template and extension one click away. The template repo is the barer, more "git-native" front door.

Lean: A. Web-Fable's recommendation, conditional on the first-screen record being live β€” which the gate enforces. Falsifier: gate morning, the record render isn't ready β†’ B salvages the slot.

15 Β· The stranger's path

Ten-minute path stops at "one seed, one question, one ruling" β€” multi-seat labeled optional?

Today's get-started assumes tmux, seats and hives from step 6 on. The stranger needs: copy template β†’ ask one ORRCF question β†’ make one ruling commit. The hive vocabulary moves to an "operators" page for the curious. Site fix is queued either way; this ratifies the on-ramp's shape.

Lean: adopt. Web-Fable's recommendation; the stopwatch claim ("ten minutes") is only honest at this scope. Falsifier: if the multi-seat topology IS the pitch for your audience β€” but the uber-brief's own pacing holds that for the pneumatic/System-3 launch.

16 Β· The launch ADR's falsifier

"Fewer than five strangers run one real decision through the template by Oct 25 β†’ the ten-minute path is wrong"?

The launch itself gets recorded as an ADR in the org with this as its falsifier β€” the campaign eating its own dogfood. The claim it tests: the path, not the audience, is what fails. Oct 25 = ten days post-launch.

Lean: adopt. It's measurable, dated, and blames the right thing. Falsifier (of the falsifier): if you'd rather measure rulings than template-copies, say the metric you'd watch instead.
V Β· Accounts & identity

17 Β· Who posts the thread

Which X account carries the launch?

HN submits as lumbroso (settled). The thread's voice is the maintainer-who-drowned story β€” a person. An org account can mirror later without costing the launch anything; creating one Thursday morning costs setup time and zero followers.

Lean: A. Single-author technical threads carry on the author's account; the brand accretes to the site and org. Falsifier: if you expect to hand the brand's voice to the community within months, C seeds it cheaply now.

18 Β· Penn in the HN first comment

Mention the affiliation, or let GitHub be the identity?

Adopt = NO mention (Web-Fable's recommendation: GitHub is the identity; the work stands on repos, not letterhead). Override = include it (academic credibility for the paper-minded reader). Note the interplay with card 8: if Penn policy is live, the affiliation question gets sharper either way.

Lean: adopt (no mention). HN reads affiliation as appeal-to-authority in a Show HN; the papers carry it where it belongs. Falsifier: if the TAS-talk audience is a launch target, one line in the HN comment does reach them.
VI Β· Two standing flips (each word is per-repo)

19 Β· muster β†’ public

Flip ADRs4AI/muster public? (This also unblocks your Anthropic classifier issue.)

State, gh-verified 10-13 00:30 ET: PRIVATE, root MIT βœ“, clean history from the 10-11 sweep (GO-ready then; reverted with the others pending your word). Your drafted Anthropic issue links muster as open source β€” it posts the moment this is true. Web-Fable's "no LICENSE yet" was stale; disregarded. The uber-brief: "mention muster in the thread only if public by post time," introduced as "the ADRs4AI CLI" (name collides with three other agent tools).

Lean: adopt (flip it). It was GO-ready in the sweep; the flip un-gates your issue post and the org pin. Falsifier: if you want muster's debut tied to its own calendar slot (Jan) instead of riding the launch, keep it private and the issue waits or drops its link.

20 Β· ios-client stays private

Confirm: iOS source stays private at launch (the moat)?

Standing recommendation since the sweep, Web-Fable concurring: the App Store link is public regardless; the source is the moat; freemium is a later docket. The thread uses a phone-ruling screenshot, not the repo. Adopt = stays private; override = flip it (would need its own sweep first β€” it has no LICENSE file yet either).

Lean: adopt (stays private). Nothing in the launch needs the source; the moat argument stands. Falsifier: a verification/App-Review need that requires public source β€” none known today.
VII Β· Lane shapes (not launch-gating)

21 Β· Bluesky delegation

You offered access β€” what shape does delegated posting take?

House law today: drafts only, your hands send, per-act consent. You proposed giving a seat Bluesky access to coordinate posting. The shapes below differ in WHERE your consent sits, never whether.

Lean: A (C's launch-week clause is already true: Thursday is your hands regardless). Which seat holds the password routes through Mint + Operator after your pick β€” not settled unilaterally. Falsifier: any discomfort with a standing credential in a seat's custody β†’ B, and the calendar just runs slower.

22 Β· The Grok lane

Grok for X intelligence: consult-lane or wired seat?

Grok-via-your-paste already produced the zeitgeist check, hook ranking, and stress-test (staked in intake). The uber-brief's Β§Prompts-for-Grok is a ready re-run before posting. A wired Grok seat (API custody, outward-law fit, cost) is a real recruitment docket.

Lean: A. It worked this week; wiring a new external-model seat during launch week multiplies surfaces at the worst time. Falsifier: reply-volume Thursday exceeding what paste-consults can triage β†’ B same week.

23 Β· The coding-projects blog

Where do the project posts live?

You asked for "a place to start blogging about all my coding projects." system3.blog is live with the Willison mechanics (note/link/essay, bin/note is the acceptance test) β€” a first post can ship this week with zero new infrastructure. ADRs4AI-practice posts stay on adrs.systems per the brand separation already ruled in the uber-brief. The release calendar (founded today in the hub) has the posting lane either way.

Lean: A. Zero infra, ships this week, the register fits project notes. Falsifier: if these posts are meant to carry YOUR name for reputation (the mission's "epistemics-first person" surface), B is the honest venue and worth the days.

Export β€” copy from here if no file was offered