Value Review

Contents

• README Positioning of the value-design repository
• Value Review Sheet L1–L4 review template
• Resources Related repository

README

vcdesign-value-review

vcdesign-value-review is a value-design repository to determine, before implementation, which value is protected inside which boundary.

This repository is neither a design manifesto nor an implementation guide. It exists to preserve traces of judgment and hand them to implementation without value loss.

What this is

• Centered on the VCDesign Value Review Sheet (L1–L4)
• Defines Δ (deltas)
• Draws the line between allowable and disallowed Δ
• Fixes gate conditions to graduate to RCL

What this is not

• Not an extension of VC-AD
• Not an implementation manual (TODO/YAML steps are on TDDD)
• Not RCL itself (RCL is produced after L4 Pass)
• Not a collection of AI constitutions or norms

Where this is effective (scope)

Suitable for projects that meet these conditions:

• Value cannot be fixed upfront (it is known only after operation)
• Failures are irreversible or high-cost (trust / downtime / data loss / hard to redo)
• Humans retain final responsibility (cannot or should not fully automate)
• Typical: infrastructure, enterprise systems, data integration, analytics/monitoring, ops automation, OT/IT boundaries

Note: For areas like games, ad landing pages, and entertainment where rapid iteration and disposal are expected, this can be too heavy.

Deliverables

• value-review-sheet.md: value-design template (L1–L4)
• delta-definition.md: minimum definition of Δ
• allowed-disallowed-delta.md: allowable / disallowed Δ line
• rcl-gate.md: conditions to graduate to RCL (a gate, not an entrance)
• scope-by-responsibility.md: optional helper for sizing by "responsibility volume"
• idk-lamp.md: optional stop lamp to visualize indecision

Minimum glossary

• Δ (delta): observable difference between design premise (P) and actual fact (D)
• Allowable Δ: recoverable while keeping human accountability
• Disallowed Δ: if left, irreversibly harms value/trust/responsibility
• RCL: a "responsibility meter" taken on by a fully maximized design (not stored in this repo)

Principle

Writing RCL before seeing the whole picture becomes blame shifting, not design.
RCL is not the entrance; it is the line of resolve taken on by a fully maximized design.

Value Review Sheet

VCDesign Value Review Sheet (v0.1)

Purpose: A review sheet to confirm value is not damaged before writing RCL.
Positioning: This sheet is a "review document". Only L4 Pass graduates to RCL.

How to use (minimum)

1. L1: List business flows (Do). Start with action rows.
2. L2: Expand PDΔA behind each Do (Man / Machine / Material)
3. L3: Select system intervention (Detect / Notify / Block / Ignore)
4. L4: Review if the whole flow maximizes value; if Pass, graduate to RCL

L1. Business flow (Do listing)

L1-1. Flow (Do) list

• Do-01:
• Do-02:
• Do-03:
• Do-04:

L1-2. Minimum boundaries

Do not use all 11 boundaries. Minimum is "People / System / Data / External".

• People (roles):
• System (target / external):
• Data (master / ledger / logs):
• External (other orgs, partners, law, physical):

L2. PDΔA per Do (Man / Machine / Material)

Do is the entrance, not the end of responsibility.
Move away from cultural PDCA and toward observable deltas.

L2 template (one block per Do)

Do-XX: <Do name>

Man (people)

• P (premise):
• D (execution):
• Δ (delta: what mismatch becomes a problem):
• A (response to Δ: who does what):
• Evidence (for explanation):

Machine (system)

• P:
• D:
• Δ:
• A:
• Evidence:

Material (data / information)

• P:
• D:
• Δ:
• A:
• Evidence:

L3. Intervention policy (what to do and how)

Decide which deltas from L2 the system will intervene on.
"Do everything" is basically NG. Not doing (Ignore) is also design.

L3-1. Intervention levels (4 choices)

• Detect: detect (make it noticeable)
• Notify: present (make it easy to execute)
• Block: stop (cannot proceed without conditions)
• Ignore: do nothing (leave to human responsibility)

L3-2. Intervention policy per Do

• Do-01: Target Δ (which in L2?): Intervention (Detect/Notify/Block/Ignore): Rationale (why there?): Never do (boundary):
• Do-02:
• Do-03:

L3-3. Only if you choose Block

• Block condition (what must be satisfied to proceed):
• Exception (emergency): who, on what basis, and how to override:
• Evidence for override (approval log / ticket / signature):

L4. Max-value review (gate)

Judge whether the shape can keep responsibility, not feature count.

L4-1. Max-value checklist (Yes/No)

• [ ] Do is not just an action list (each Do has Δ)
• [ ] Disallowed Δ is clear (not left ambiguous as "normal")
• [ ] Intervention is not excessive (Ignore is intentionally left)
• [ ] Automation does not steal judgment (accountability remains human)
• [ ] It can explain who looked at what in failure (Evidence exists)
• [ ] Boundaries are not crossed (decisions not to be made were not made)

L4-2. Conclusion (short)

• Value maximized by this design:
• Value intentionally dropped:
• Top 3 risks that damage max value:

L4-3. Judgment

• Decision: Pass / Revise / Fail
• Pass conditions (to graduate to RCL):
• Revise points (which of L1/L2/L3 to redo):

Graduate to RCL (outside this sheet)

Only L4 Pass creates an RCL sheet.
RCL is not a tool to build, but a meter to protect design value.

Resources

• vcdesign-value-review (GitHub) Value Review repository