198. Delivery is judged by verified outcomes landed per day, with lead time p95 as its control, never by commits, pull requests or sessions
Context
Audits of 2026-10-14..18 found that 52 percent of the last 100 commits were merges, that four agents opening four pull requests did not do four times the work, and proposed several primary metrics: verified deliveries merged per 24 hours, lead time p95, autonomous intents completed, time from finished intent to user-available function without regression. The milestone cost-per-accepted-outcome measures cost only on measured data.
Decision
The repository's delivery figure is the count of verified outcomes (an issue DONE with its evidence, or a release published and observed) per rolling day, read with the per-phase lead time p95 of I3203 as its control. Commit, pull request and session counts are activity, may be shown as such, and are never called throughput.
Alternatives rejected
- Commits or merged pull requests per day as throughput
- One composite score (opaque, unfalsifiable)
- Forecast maturity percentages
Consequences
No issue in this plan implements the figure yet: I3203 measures lead time only, and an issue that counts verified outcomes is planned once the owner accepts the definition. The drain and dashboard label activity as activity; an outcome without evidence does not count.
Sources
Promoted on 2026-10-18 from the owner's ChatGPT project (workspace chatgpt-majordomus, issue I3100). The owner has not accepted it; it is proposed for the owner's decision.
chatgpt-majordomus:c259#6chatgpt-majordomus:c271#9chatgpt-majordomus:c287#2chatgpt-majordomus:c295#7chatgpt-majordomus:c295#9chatgpt-majordomus:c297#6