Files
claude 9124275bf3 fieldtest: m1-self-description — 5 examples, 1 bug / 3 friction / 3 spec-gap
Binary-only milestone fieldtest for milestone 36 (self-description): a
fieldtester agent bootstrapped from the release binary (d26f0c8) alone —
no engine sources, no repo docs, no bootstrap card — and drove five
end-to-end scenarios: cold bootstrap -> author -> validate (green,
synthetic + real EURUSD, 3944 trades), execution semantics +
latch/edge-pulse idiom (green), vocabulary breadth (33 nodes / 7 folds /
17 metrics / 6+6 blocks, every entry with a meaning), document ramp from
a bare {} (green, 3 members), trace/measurement path (dead-ends,
tracked).

Verdict: the bootstrap card (#267, closed as superseded) is empirically
redundant for the bootstrap->author->validate spine. Two seams keep the
promise short of end-to-end, both tracked outside M1: the declared-tap
drop on the registration path (#327, the missing mechanism of #312/M6)
and the sugar<->document axis namespace (#328, prerequisite of the #319
retirement). Friction routed: #329 #330 #331; evidence comments on #312
and #324. Milestone 36 closed with this adjudication; the fieldtest spec
remains the usual git-ignored working file.

refs #267, refs #312, refs #319
2026-07-24 11:06:22 +02:00
..

s1 — cold bootstrap → author → validate

Protected property: starting from aura --help alone (no repo, no docs), a downstream agent can scaffold a project, author a new strategy over the discovered op-grammar vocabulary, and drive it to a green backtest — using only what the binary emits.

Trail (binary-only)

  1. aura --help — states what aura is, the two-layer model, and the execution model (bias in [-1,+1], held as target position, stop = R).
  2. aura new demo-strat — scaffolds Aura.toml, blueprints/signal.json, and a machine-written CLAUDE.md bootstrap card.
  3. aura graph build --help — the op-list authoring grammar (one line of meaning per op).
  4. aura graph introspect --vocabulary / --node <T> — node types and port schemas.
  5. Author ema_cross.oplist.json (this dir); build it: aura graph build < ema_cross.oplist.json > blueprints/ema_cross.json
  6. Validate green:
    • synthetic: aura run blueprints/ema_cross.json → exit 0 (18-cycle synthetic stream → 0 trades; all-zero metrics, no note that the stream is tiny).
    • real: aura run blueprints/ema_cross.json --real EURUSD --from 1577836800000 --to 1583020800000 → 3944 trades, valid metrics JSON, exit 0.

ema_cross.blueprint.json is the built artefact graph build emitted.

Notes

  • The harness auto-wraps the bias with sim-optimal+risk-executor and reports R metrics even though the blueprint emits only a bias — confirming the documented execution model without any extra authoring.
  • Minor friction: graph build has no op to set the blueprint name; it defaults to graph, and that name leaks into the sweep axis namespace (graph.fast.length).