Skip to content

Close cycle 02 — two accepted, one never started, and neither criterion moved - #27

Merged
napkinstack-agent[bot] merged 1 commit into
mainfrom
cycle/02-closure
Sep 23, 2026
Merged

napkinstack-agent[bot] merged 1 commit into
mainfrom
cycle/02-closure

Conversation

@napkinstack-agent

Copy link
Copy Markdown
Contributor

Close cycle 02 — two accepted, one never started, and neither criterion moved


Module

None. This pull request changes no module: docs/project/ only.

  • This pull request touches a single module — n/a, it touches none
  • ⚠️ Cross-module

Deliverable

Out of cycle: closing the cycle. A pull request that changes no module beyond its description is
not delivery work, and this one closes the cycle itself.

What it changes

Cycle 02 is closed — closed / shipped, ended 2026-09-23, on day 3 of a 21-day appetite.
D3's state becomes accepted, joining D2. D1 stays deferred.

shipped rather than completed: completed would claim every deliverable was accepted, and D1
was not. Nor is this the circuit breaker — the end date was 19 days away. The cycle closed early
because everything it could do with engineering was done, and what is left is not engineering.

The closure says four things it owes:

  1. What was delivered. D2, the window read that ends at the moment of the run (D2 — read the window backwards, so what a spent budget drops is the past #20). D3, three
    verdicts with the exit codes ordered by severity and a ceiling that refuses to answer past 30×
    (D3 — three verdicts, so the headline cannot say yes to a trader the answer is warning about #26). And, outside the deliverables, ADR-0002, ADR-0003, PDR-0002 with four amendments,
    PDR-0003 and handover.md.
  2. Which success criteria moved: neither. All five of the charter's phase-1 criteria need the
    copying module, which does not exist. The fifth — screening's predictions confronted with
    reality — is the only one this cycle touched, and only by making it possible: the module now
    states a fee burden, a refused share and a reproducible share precisely enough to be wrong.
    Nothing has confronted them with a member's fills, because no member has any.
  3. The sentence promised when the cycle was framed, plus the half nobody predicted: the project
    built a tool for choosing traders with no evidence that any member wants to copy one, and that
    was a decision rather than an oversight. It was offered one week on D1 alone and chose three
    weeks on everything. It did not repeat cycle 01 — both engineering deliverables landed, in
    three days. It repeated the other half: the deliverable that was somebody else's calendar did
    not move, and 19 days of appetite are unspent because the appetite was sized for work cycle 01
    had already done.
  4. What the cycle cost, measured. Six adversarial rounds and six confirmation runs on one
    read-only module: eleven real defects, five refusals, every refusal correct. Three defects the
    author introduced in later rounds inside changes nobody asked for; three refusals over a
    sentence in a decision record rather than over code.
    With the transferable lesson, which
    belongs to cycle 03: a class-level statement ends the sequence, a per-instance clause produces
    one refusal per round.

The charter is corrected, not changed

PDR-0003 was merged in #24 with the charter's own "Success criteria" section left as it was. It
still stated the twenty-members-at-day-30 count that the PDR replaced, while the open questions
three screens below announced the revision as done. A closure that must say which criteria moved
cannot honestly be written against a section the decider had already superseded.

So the five phase-1 pilot criteria are now written where the criteria live — faithful copying, zero
key incident, revocation exercised, 60% continuing past day 14, and screening's predictions
confronted with reality — with the arithmetic that removed the count: five invitations return
refuted 47% of the time while the product is exactly as good as forecast. The commission
criterion is where PDR-0003 put it, behind the phase-2 gate.

This applies an accepted decision. It makes no new one, and the charter still changes only through
a PDR (docs/os/06-decisions.md §4).

One thing is owed by the decider, and I have not invented it. PDR-0003 moved the commission
criterion to phase 2 "with a later date" and named no date. The old before 2027-03-21 is not
restated as though it had survived. It is recorded in the charter's open questions, because a
viability criterion without a deadline is a criterion that cannot fail.

handover.md

Rewritten for the next session: nothing in flight, main at 8277dca, cycle 03 to frame around
copying, the framework upgrade owed at this boundary, and the operational lessons that cost this
cycle time — a push dismisses an approval while a description edit does not; T4 makes any commit,
including a documentation-only one, reprint the whole test sheet; a delegated verifier has no
session identifier of its own.

It also corrects a mistake made while writing it: the hook citing "ADR-0004" cites NapkinStack's
ADR-0004, which this project's docs/adr/0001 names and explains. It is not a missing decision, and
the handover now says so rather than sending the next session looking for it.


Oracle

  • The success criterion was written before the implementation — the closure's required
    contents are the template's four questions, and the cycle's framing promised one sentence in
    advance
  • It was seen failing, for the right reason — nstack fitness refuses a closed cycle with
    no outcome and no ended_on (C6), and refuses an unknown deliverable state (C4)
  • It passes now — nstack fitness exit 0, plan compliant

Kind of oracle: fitness function


Review budget

  • Lines changed (excluding generated): well within budget
  • Files touched: 3
  • Within budget

Contracts

  • No contract affected

Definition of Done

  • Oracle green — nstack fitness exit 0, read on its own exit status
  • Lint, format, types — n/a, no code; nstack check screening still 0 on this branch
  • Unit tests — n/a, no code touched; 81 still pass on main
  • Fitness functions — compliant
  • Contract tests, build, accessibility, observability, rollback — n/a
  • Affected documentation updated — this pull request is the documentation
  • Diff self-reviewed line by line

Test sheet

Not required: this pull request changes no module beyond its description
(docs/os/05-workflow.md §7).


Summary

DONE
Cycle 02 closed, shipped, with its closure written to the template's four questions. D3 marked
accepted. The charter's success criteria brought into line with PDR-0003, which #24 had left
half-applied. handover.md rewritten for the next session.

VERIFIED
nstack fitness exit 0 and plan compliant, read on its own exit status. The closure's figures come
from the merged pull requests and from the confirmation runs recorded in #26, not from memory.

ASSUMED
That closing early is what the decider wants, rather than spending the remaining 19 days on D1
inside this cycle. The alternative was available and is named in the closure; closing is what lets
cycle 03 be accepted, since only one cycle may be accepted at a time.

NOT VERIFIED
Nothing in this pull request is testable beyond the fitness function. Whether the closure's reading
of the cycle is right is the decider's call, which is what this review is.

RISKS
The closure is blunt about the project's position on purpose. If it is wrong, it is wrong in
writing and can be argued with — which is better than the alternative. And the phase-2 date stays
open: the project currently has a viability criterion that cannot fail.


Signals to report

  • I had to look at another module's implementation
  • A kernel or playbook rule got in my way for no good reason
  • A quality gate blocked without improving quality
  • A rule applied by hand would be worth automating

The signal. PDR-0003 revised a success criterion, and the charter's criteria section kept the
old one for a day, with the same file announcing the revision three screens below. Nothing
mechanical noticed. nstack fitness already reads both files; a PDR that says it replaces a
criterion could be required to name which one, and the criterion could be required to name the PDR
that last changed it. Otherwise the only thing standing between an accepted decision and a stale
charter is somebody reading both on the same afternoon.

…on moved

`closed` / `shipped`, ended 2026-09-23, on day 3 of a 21-day appetite. Not
`completed`, because that would claim every deliverable was accepted and D1 was
not; not the circuit breaker either, because the end date was 19 days away.

D2 and D3 accepted (#20, #26). D1 — five private invitations — deferred for the
second cycle running: nobody has been asked, and the cost is five messages and
fourteen days of waiting.

The closure records the sentence it owed in advance, and one it did not: both
engineering deliverables landed in three days, so this cycle did not repeat cycle
01's shape — it repeated the other half. The deliverable that was somebody else's
calendar did not move, and 19 days of appetite are unspent because the appetite
was sized for work cycle 01 had already done.

It also records what the cycle cost, measured: six adversarial rounds and six
confirmation runs on one read-only module, eleven real defects, five refusals.
Three defects the author introduced in later rounds; three refusals over a
sentence in a decision record rather than over code. And the transferable lesson,
which belongs to cycle 03 rather than to this closure: a class-level statement
ends the sequence, a per-instance clause produces one refusal per round.

**The charter is corrected, not changed.** PDR-0003 was merged in #24 with the
charter's own "Success criteria" section left as it was: it still stated the
twenty-members-at-day-30 count that the PDR replaced, while the open questions
three screens below announced the revision. A closure that must say which
criteria moved cannot be written against a section the decider had already
superseded. The five phase-1 pilot criteria are now written where the criteria
live, with the arithmetic that removed the count, and the commission criterion is
where PDR-0003 put it — behind the phase-2 gate.

One thing came out of that and belongs to the decider: PDR-0003 moved the
commission criterion "with a later date" and **named no date**. I have not
invented one. It is recorded in the charter's open questions as owed, because a
viability criterion without a deadline is a criterion that cannot fail.

`handover.md` is rewritten for the next session: nothing in flight, cycle 03 to
frame around `copying`, the framework upgrade owed at this boundary, and the
operational lessons that cost this cycle time — a push dismisses an approval
while a description edit does not, T4 makes any commit reprint the whole sheet,
and a delegated verifier has no session identifier of its own.

`nstack fitness` compliant, plan included.

Agent-Session: 4151570f-613e-475f-ba3f-b465b35d9a83
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@napkinstack-agent napkinstack-agent Bot mentioned this pull request Sep 23, 2026
12 of 17 tasks
@napkinstack-agent
napkinstack-agent Bot merged commit 5f7bc99 into main Sep 23, 2026
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants