Historical run

View current dashboard →

Overall trajectory · last 28 runs

20260421T185438Z_f2f7facompleted_with_errors•Apr 21, 2026, 06:54 PM UTC•9/12 executions · 3 failed•rubric v2

gemini / gemini-2.5-flash

Overall

42.2+1.9

weighted mean across 5 dimensions

Visibility

55.6+5.6

first-mention position of ISL

Module recognition

22.2-2.8

WG/FG named = 100, DRx only = 40

Descriptor match

11.1+3.2

rubric vocabulary match, not factual accuracy

Overall mix35Visibility·25Module recognition·20Descriptor match·15Competitor displacement·5Citation quality

Provider comparison — Overall score

Same data as the “By provider” table below, visualized. Each bar is the mean Overall score across that provider’s substantive responses (evasive/empty excluded).

Notable changes vs prior run

GAINaverages_substantive.visibility50.0→55.6+5.6
GAINaverages_substantive.accuracy7.9→11.1+3.2
GAINcitations.no_citations12.0→9.0-3.0
MISScompetitor_displacement.no_competitors12.0→9.0-3.0
MISScounts.substantive12.0→9.0-3.0

By provider

Averages across substantive responses for each AI provider. n = executions attempted; n sub = responses that actually answered (evasive/empty/unscorable excluded). Low n sub triggers an amber flag — the average is over too few real responses to trust.

Provider	n	n sub	Overall	Vis	ModRec	Desc	CompDisp	Citation
gemini	9	9	42.2	55.6	22.2	11.1	100.0	0.0

By category

Same averages as 'By provider', but grouped by prompt category instead. Useful for spotting which capability area (e.g., WeatherGuard, ISL company) is scoring strongest across the provider panel.

Category	n	n sub	Overall	Vis	ModRec	Desc	CompDisp	Citation
isl_company	5	5	50.8	100.0	0.0	4.0	100.0	0.0
weatherguardlow n	4	4	31.5	0.0	50.0	20.0	100.0	0.0

ISL prominence

prominent5
absent4

Module recognition

none7
specific_module_named2
weatherguard_mentioned_total2

Competitor displacement

Competitors tracked: Urbint, Gridware, Space-Time Insight.

no_competitors9

Citations

“Authoritative” = our 7-domain whitelist, not a broad quality judgement.

no_citations9

Dimension averages (substantive)

Visibility55.6

Descriptor match11.1

Module recognition22.2

Competitor displacement100.0

Citation quality0.0

Prompt library used in this run

The exact questions sent to each AI provider. Authoritative source: prompts/*.yaml.

20 prompts · 60 variants

ev_prophet

fireguard

isl_company

reliability_optimization

sos

weatherguard

Executions9 of 9

One row per provider × prompt × variant. Click any row to expand the raw response and parsed signals. Scores are for this single execution, not averages.

ProviderCategory

Provider	Prompt	Category	Class	Overall	Vis	ModRec	Desc	CompDisp
gemini	isl_002/v2	isl_company	substantive	54.0	100.0	0.0	20.0	100.0
gemini	isl_001/v3	isl_company	substantive	50.0	100.0	0.0	0.0	100.0
gemini	isl_002/v3	isl_company	substantive	50.0	100.0	0.0	0.0	100.0
gemini	isl_001/v2	isl_company	substantive	50.0	100.0	0.0	0.0	100.0
gemini	isl_002/v1	isl_company	substantive	50.0	100.0	0.0	0.0	100.0
gemini	wg_002/v3	weatherguard	substantive	48.0	0.0	100.0	40.0	100.0
gemini	wg_002/v1	weatherguard	substantive	40.0	0.0	100.0	0.0	100.0
gemini	wg_001/v2	weatherguard	substantive	19.0	0.0	0.0	20.0	100.0
gemini	wg_001/v3	weatherguard	substantive	19.0	0.0	0.0	20.0	100.0