| artifact_key | writers (agent · writes) | readers (agent · reads) | versions |
|---|---|---|---|
{{ row.artifact_key }}
{% if row.append_only %}append-only{% endif %}
|
{% if row.writers %} {% for agent, count in row.writers %}{{ agent }} ({{ count }}){% if not loop.last %}, {% endif %}{% endfor %} {% else %} none observed {% endif %} | {% if row.readers %} {% for agent, count in row.readers %}{{ agent }} ({{ count }}){% if not loop.last %}, {% endif %}{% endfor %} {% else %} none observed {% endif %} | {{ row.version_range }} |
{{ headline_subtitle }}
{% if is_insufficient %}Run was too short to classify
{{ verdict_reason }}
Rework cost: unmeasurable
{{ cost_unmeasurable_message }}
Rework cost (annualized): {{ rework_cost_annualized_str }}
floor — does not include broadcast rebroadcasting or redundant fetches. Based on {{ rework_tokens_this_run }} tokens this run.
{{ headline_event_count }} divergence event{{ '' if headline_event_count == 1 else 's' }} affecting {{ agent_pain_count }} node{{ '' if agent_pain_count == 1 else 's' }}
agent_pain_count: distinct later_read.node values across headline events.
Forensic mini-timeline of the highest-impact divergence event. Field names follow ReadObservation semantics — runtime "handed" the node a value at this version, not "node read" attribution.
artifact_key: {{ top_event.artifact_key }}
earlier_read: node={{ top_event.earlier_read.node }} tick={{ top_event.earlier_read.tick }} version={{ top_event.earlier_read.version }}
later_read: node={{ top_event.later_read.node }} tick={{ top_event.later_read.tick }} version={{ top_event.later_read.version }}
{% if top_event.canonical_writer %}canonical_writer: {{ top_event.canonical_writer }} (tick {{ top_event.canonical_writer_tick }}) — most recent observed writer at later_read.tick
{% endif %} {% if top_event.is_sequential_staleness %}classification: sequential_staleness (promoted to headline by --strict)
{% endif %}No headline divergence events were observed. The map below shows every tracked artifact's writers and readers as observed during the run. Multi-writer rows are highlighted — these are the shared_artifact shape at the artifact level even when no individual divergence event was caught.
{{ render_ownership_table(ownership) }} {% endif %} {# ============================================================ #} {# Section 3 - Per-Artifact Heatmap #} {# ============================================================ #} {% if show_heatmap %}
divergent_reads is the count of distinct reads handed a version
that diverged from another read of the same artifact; share is
that count over total_reads. Witness-quality — a divergent
handing is observed; whether the node acted on the value is unobservable.
Rows are ordered with genuine multi-writer artifacts first
— those are the coordination signal. A high share on a single-writer
artifact is usually expected pipeline ordering (readers handed the pre-write
value), not a coordination gap. The Event That Matters Most panel above
ranks by rework impact (highest divergent_reads); this table ranks
by coordination signal (multi-writer first), so the two panels may highlight
different artifacts.
| artifact_key | writers | divergent_reads | total_reads | share |
|---|---|---|---|---|
{{ row.artifact_key }} |
{% if row.writer_count > 0 %}{{ row.writer_count }}{% else %}—{% endif %} {% if row.is_multi_writer %}multi-writer{% elif row.writer_count == 1 %}pipeline ordering{% endif %} | {{ row.divergent_reads }} | {{ row.total_reads }} | {% set pct = (row.divergent_reads * 100 / row.total_reads) if row.total_reads > 0 else 0 %} {% set bar_pct = 100 if pct > 100 else pct %} {{ pct|round(1) }}% |
Top (earlier_read.node, later_read.node) pairs by headline divergence event count. Pair semantics are observation-only — both nodes were *handed* the artifact at different versions; whether they read it is unobservable.
| earlier_reader | later_reader | events |
|---|---|---|
| {{ pair.earlier_reader }} | {{ pair.later_reader }} | {{ pair.event_count }} |
Re-run with --strict to promote sequential_staleness
exclusions back into the headline count.
Tracked: {{ tracked_keys|length }} · Ignored: {{ ignored_framework_keys|length + ignored_ephemera_keys|length }} ({{ ignored_framework_keys|length }} framework, {{ ignored_ephemera_keys|length }} ephemera) · Append-only: {{ append_only_keys|length }} · Mutable: {{ mutable_keys|length }}
{% if tracked_keys %}{% for key in tracked_keys %}{{ key }}{% if not loop.last %} {% endif %}{% endfor %}
{% for key in append_only_keys %}{{ key }}append-only{% if not loop.last %} {% endif %}{% endfor %}
{% for key in ignored_framework_keys %}{{ key }}{% if not loop.last %} {% endif %}{% endfor %}
{% for key in ignored_ephemera_keys %}{{ key }}{% if not loop.last %} {% endif %}{% endfor %}
__-prefixed
key{{ '' if unknown_underscore_keys|length == 1 else 's' }} not in the
v0-preview framework allow-list. v0 ignores them safely; the
calibration corpus tracks them so future versions add explicit handling.
Detected:
{% for key in unknown_underscore_keys %}{{ key }}{% if not loop.last %} {% endif %}{% endfor %}
{{ coverage.tick_count }} turns · {{ coverage.read_count }} reads · {{ coverage.write_count }} writes · {{ coverage.artifact_count }} artifacts · {{ confidence_label }} confidence
Thresholds for high confidence: tick_count >= {{ coverage_thresholds.tick_count }}, read_count >= {{ coverage_thresholds.read_count }}, write_count >= {{ coverage_thresholds.write_count }}.
For our call, two questions I'd love your read on:
30-min walk-through — I'll trace your top-3 divergence events live.
{{ soft_ask_message }}
Or reply to {{ contact_email }}.
{% elif cta_variant == "forward_looking" %}Zero divergence events in this run. If any of these are on your roadmap, the conversation is worth having before you build:
{{ soft_ask_message }}
{% else %} {# insufficient #}Run was too short to classify — coverage shape: {{ coverage.tick_count }} ticks / {{ coverage.read_count }} reads / {{ coverage.write_count }} writes. Re-run on a longer workload, or schedule a 15-min call to discuss what shape would tell us something: {{ book_a_call_url }}.
{% endif %}