MyAIDevelopment · Product · Operations notes

A Shipping System

Learning goal: Know why prototypes fail to ship, and build a shipping system made of a definition of done, scope cutting, a repeatable launch checklist, a launch cadence and a 30-day post-launch routine.

AI agents made building faster, but shipping does not get faster on its own. That is why the example studio (assumption) has twelve prototypes and one launch. Shipping is made by a system, not by willpower.

As of: 2026-09-29.

Key Concepts

Five Reasons Prototypes Never Ship

Cause Symptom How a system prevents it
Scope creep "Just one more thing" repeats Fix the time first and cut scope (appetite)
Polish loop You keep reworking screens that already work Write the bar for "good enough" into the definition of done
Fear of launch The launch date keeps slipping Ship small and often to lighten each launch
No distribution Only after building: "who do I tell?" Decide where the first 100 users come from before building
Context switching Touching twelve projects a little in turn Limit work in progress (WIP) (document 02)

A Definition of Done for a Solo Studio

Shippable = all of the following are true.

Principles

Fix the Time and Cut the Scope

Ship Small and Often

flowchart LR
    I[Idea] --> A[Set the appetite]
    A --> M[Cut scope with MoSCoW]
    M --> B[Build]
    B --> D{Meets definition of done}
    D -->|no| M
    D -->|yes| L[Launch checklist]
    L --> R[Launch]
    R --> F[30-day routine]
    F --> G[Gate review in document 09]

Applied: The Example Studio

The example studio gives the MVP of T1, a pixel art editor (creative tool), a 60-hour appetite (assumption). At 16 hours a week on T1 out of 160 hours a month, that is about four weeks.

Class Items Hours Share
Must Canvas, palette, two layers, PNG export, payment (early bird) 36 hours 60%
Should Four animation frames, shortcuts 12 hours 20%
Could Themes, share link 12 hours 20%
Won't (not this time) Collaboration, mobile app, plugins 0 hours 0%
Total 60 hours 100%

If Must exceeds 60%, cut the scope again. If the Musts are not done at hour 45, drop the Coulds and shrink the Shoulds. The launch date does not move.

Launch Cadence

A Repeatable Launch Checklist

[D-14] Confirm three sources of the first 100 users, publish a waitlist page
[D-7]  Check the definition of done, run one test payment and one refund
[D-3]  Finalize store and landing copy, screenshots, price table
[D-1]  Verify measurement events, verify the rollback path, three announcement drafts
[D-0]  Deploy, email the waitlist, announce in three channels, answer questions for the first 24 hours

Design the first 100 users before the product. For choosing channels and tiering launches, see the Marketing track's "Go-to-Market · Launch · PLG and SLG" and "Channel Strategy · Content · Paid · Lifecycle · Community".

The 30-Day Post-Launch Routine

When What to do Number to watch
D+1 Error logs, payment failures, answer first questions Critical error count
D+3 Fix where users get stuck in activation Sign-up to core task completion rate
D+7 Short conversations with five users (the Mom Test from document 03) First-week retention, first payments
D+14 One price or onboarding experiment Conversion rate
D+30 Gate review in document 09: kill, continue or double down Thresholds set in advance

Bad example:

Started a new prototype the day after launch. A month later, the payment button turned out to have been broken on mobile.

Good example:

Put the 30-day routine on the calendar first, and set new-prototype time to zero for that period.

Going Deeper

Running the Release Pipeline with AI Agents

Stage Work for the agent Gate a human keeps
Build and test Run tests, lint and builds in CI, summarize failures What the tests actually verify
Release notes Draft changes from commits and issues The sentences sent to users
Store assets Fit screenshot specs, draft descriptions and translations Whether promised features really exist
Deployment Deploy to staging, run smoke tests Production approval and rollback decisions
After launch Summarize error logs, triage questions Answers on refunds and personal data

Build the pipeline following the Platform track's "CI/CD Pipeline: From Commit to Production" and "Practical Guide for Small Teams: Solo Developers and Small Teams". Keep payments, personal data and production approval as human gates.

Common Misconceptions

Self-Check Questions

  1. Pick which of the five causes kept one of your prototypes from shipping, and write down the system that would prevent it.
  2. With an 80-hour appetite, what is the most time Must can take under DSDM guidance?
  3. How does an appetite differ from an estimate?
  4. If the example studio used a 5-week cycle (4 weeks + 1-week cooldown) instead of the 8-week cycle (6 weeks + 2-week cooldown), how many launch chances would it have in a year? Considering the 30-day post-launch routine and the cooldown work (bug fixes, gate reviews, the next bet), argue which one fits this studio.
  5. Why should you not start a new prototype during the 30 days after launch?

References