template · model update

Model update template

Copy-paste skeleton. Written the same hour as the retro. Four moves: close witnessed assumptions, add new ones, append signal reading to the brief, sharpen open questions. The artefact the next cycle inherits.

How to use

The model update is mechanical. Four moves, in order. Open the Initiative Brief and the Feature Brief side-by-side. The update is in-place edits to those briefs — not a new document. This template captures the change record — what moved, with date and evidence.

text

# Model update — [cycle name] · YYYY-MM-DD

Author:    [PO name]
Triggered: Retro held YYYY-MM-DD; signal reading YYYY-MM-DD

## 1. Assumptions closed (witnessed during this cycle)

  Brief / Initiative:  [link]
  Assumption:          "[verbatim from the brief]"
  Status:              CONFIRMED | CONTRADICTED
  Evidence:            [how the cycle witnessed it]
  Closed:              YYYY-MM-DD

  [Repeat per closed assumption]

## 2. Assumptions added (not yet witnessed at decision time)

  New assumption:      "[what the cycle surfaced that we hadn't
                        considered]"
  Status:              NOT WITNESSED at decision time, OBSERVED
                       during check
  Implication:         [what this means for the next cycle's
                        scope or design]
  Added to:            [which brief, which section]

  [Repeat per added assumption]

## 3. Signal reading appended

  Brief:           [Feature Brief link]
  Signal reading:  [link]
  Method:          [Appended as new section in the brief, not
                    copied — the brief now points at the result
                    alongside the prediction.]

## 4. Open questions sharpened

  Previous question:  "[old phrasing]"
  Resolved:           [strike — answered during the cycle]
  - or -
  New phrasing:       "[sharper question for the next cycle]"
  Why sharpened:      [what this cycle taught us about what to ask]
  - or -
  New question:       "[surfaced during the cycle]"

  [Repeat per question]

## 5. Templates / checklists / glossary updated

  Artefact:    [Feature Brief template | amigos template |
                release-gate checklist | domain glossary | etc.]
  Change:      [What was added/removed/sharpened]
  Why:         [Link to the cycle finding that motivated it]
  PR:          [link if relevant]

  [Repeat per artefact touched]

## Sign-off
PO: [Name] · [date]

The test for a real model update

If the next cycle's brief does not change because of what this cycle revealed, the update did not happen. A wiki entry that lives in isolation is information; a model update is a change to the artefacts the next cycle uses without anyone remembering to look.

Where this lives in your project

Model updates are change records filed alongside the Initiative Brief — typically in briefs/<initiative>/model-updates/. They reference the in-place edits to the briefs themselves (which are versioned in git). The next cycle's discovery reads the updated Initiative Brief; the model update record is the audit trail.

What to do if a section resists

Resistance	What it means	Where to go
Cannot name a closed assumption	The cycle did not witness anything new	Either the cycle ran without observation (chain-level signal) or the brief had no assumptions to close (also a signal)
No new assumptions surfaced	Either the cycle was over-confident or under-witnessed	Often the latter — re-read the helpdesk reading and signal reading for what the team did not expect
Template change is vague	"Improve the template" — not testable	Name the specific section, the specific addition, the specific check
Update done a week later	Loss already happening — momentum lost	The discipline is same hour as the retro; this slipped because no one owns it. Assign explicitly.

Worked example — Grading-shortcut model update

text

# Model update — Grading shortcut · 2026-06-27

Author:    Alex (PO)
Triggered: Retro held 2026-06-27 ; signal reading 2026-06-25

## 1. Assumptions closed

  Brief:        Feature Brief — Grading shortcut
  Assumption:   "Graders prefer keyboard shortcuts over deep
                links."
  Status:       CONFIRMED
  Evidence:     8 of 8 graders observed using ⌘+Enter to advance.
                Deep links used by 1 grader once across 8 sessions.
  Closed:       2026-06-25

## 2. Assumptions added

  New assumption:  "Graders working in non-English language packs
                    experience a perceived-freeze at submit due
                    to language-pack load delay."
  Status:          OBSERVED during check (3 schools, 8 tickets)
  Implication:    Next cycle needs to witness a Hebrew-default
                   grader's submit moment. Likely brief input
                   for a perceived-responsiveness initiative.
  Added to:        Initiative Brief — Grading · "Not yet witnessed"

## 3. Signal reading appended

  Brief:           /briefs/grading-shortcut.md
  Signal reading:  /briefs/grading-shortcut/signal-2026-06-25.md
  Method:          Linked from "Prediction" section as
                   "Result (2026-06-25): Met".

## 4. Open questions sharpened

  Previous question:  "Will graders use the new shortcut?"
  Resolved:           YES — 8/8 adoption.

  New question:       "What is the median submit time experienced
                       by a non-English-default grader?"
  Why surfaced:        Signal reading + helpdesk pattern revealed
                       a language-pack load delay we hadn't witnessed.

## 5. Templates / checklists updated

  Artefact:   Feature Brief template — "What we have NOT yet
               witnessed" section
  Change:     Add required prompt: "Have we witnessed a
               non-English-default user in this moment?"
  Why:        This cycle surfaced an observation-mismatch
               specific to non-English-default graders.
  PR:         #248 (templates/feature-brief.md)

PO: Alex · 2026-06-27

Model update template ​

The test for a real model update ​

Where this lives in your project ​

What to do if a section resists ​

Worked example — Grading-shortcut model update ​

See also ​

Model update template

The test for a real model update

Where this lives in your project

What to do if a section resists

Worked example — Grading-shortcut model update

See also