I3298 — The benchmark declares one arm per mechanism and per agent topology, and a short-task hypothesis beside the long-work one

○READY wave 0 · p3 · implementation profile · parallel safe

Part of every-sessions-usage-is-observed — What every AI session consumed is observed from the provider, attributed to the work it served, and never estimated into a fact.

Objective

Extend .ai/repo/benchmarks/economics/methodology.yaml with treatment arms that isolate context compilation, handover, skills and single-agent versus architect-plus-workers, and record the hypothesis that a short isolated task gains little or loses, so a later total can be explained by parts.

Why

The only declared hypothesis is long-running-work (30-70 percent, untested); the owner asked whether one agent or fan-out is cheaper, and the repository's answer is measurement through ADR 0082, not a second economics subsystem. On 2026-10-18 `majordomus economics summary` (master 707a04c) reports 0 valid pairs of the 30 the rule asks for. It is ADR 0082 benchmark design, so it serves the intent's savings-are-measured-by-arm, the proof of the statement's 'saves', rather than no-estimate-shown-as-spend-or-saving; it stays bound to this intent because milestone cost-per-accepted-outcome belongs to no intent. It extends milestone cost-per-accepted-outcome. Sources: chatgpt-majordomus:c162#2, chatgpt-majordomus:c162#4, chatgpt-majordomus:c213#2, chatgpt-majordomus:c252#5, chatgpt-majordomus:c162#3, chatgpt-majordomus:c231#3, chatgpt-majordomus:c046#5, chatgpt-majordomus:c286#12, chatgpt-majordomus:c278#4.

Current state

`majordomus economics summary` on master 707a04c (2026-10-18): suite context 9 runs, suite pilot 0 runs with 14 pairs declared; one hypothesis.

Desired state

Arms and hypotheses declared as data, validated, and listed by economics summary as unmeasured until run.

Scope

• apps/majordomus-cli/src/economics
• docs/ECONOMICS.md
• .ai/repo/benchmarks/economics/methodology.yaml
• .ai/repo/benchmarks/economics/README.md
• apps/majordomus-cli/tests/economics.rs
• docs/generated/economics.json
• docs/generated/economics.md

Out of scope

• Running the live suite (spends provider usage; owner decision)

Dependencies

None. This issue is a root of the graph.

What waits on this

• I3386✕BLOCKED The economics runner runs one job as several concurrent workers in worktrees of one workspace sharing the treatment's server, and records the job's wall time to an integrated, accepted tree

Acceptance criteria

• economics summary lists every arm and hypothesis with status untested
• An arm without a declared baseline is refused

Validation

• majordomus economics summary
• cargo test --manifest-path apps/majordomus-cli/Cargo.toml --test economics

Evidence required

• case_mechanism_arms_are_declared

Evidence

None recorded. Every token above needs a command or an artifact behind it before this issue can be completed; narrative is refused.

Timeline

started

—

verified

—

completed

—

Those three fields, the evidence above and the state of the dependencies are all the status is made of. There is no status field to disagree with them.

Canonical record: .ai/repo/project/issues/I3298.yaml. Read it back with majordomus plan show I3298.