Agentic Resource Radar

Methodology

One pinned reference agent runs every task against every provider identically (bounded loop, max 8 search calls per task); LLM judges grade the answers and are themselves audited against human labels before anything publishes. Nothing is blended into one number: the Agentic Search Index is a pure quality construct, and cost and latency are separate published axes.

Judge Calibration

Two independent humans grading the same answers agreed 95.8% of the time (Cohen's kappa 0.917, n=48). That is the ceiling any judge can reach. Our judge matched the human consensus 94.1% of the time (kappa 0.85, n=34). A second judge model agreed with the pinned judge's pass-fail call 95.1% of the time across 122 overlapping cases. 3,267 runs were executed and 3,263 verdicts were scored after voiding harness-fault runs.

Judge vs human consensus
94.1%
kappa 0.85 · n=34
Human-human ceiling
95.8%
kappa 0.917 · n=48
Task bank
121 tasks × 3
3 repeats per task
Scored verdicts
3,263
9 providers
ComparisonnAgreementKappa
Human vs human (ceiling)4895.8%0.917
Judge vs human consensus3494.1%0.85
Judge vs second judge model12295.1% pass-failn/a

Task Bank

KindTasksWhat it tests
Static Facts45Short factual questions with one indisputable answer.
Dynamic Facts28Questions whose answers change with the world.
Multi-Source40Questions requiring integration of multiple sources.
Deep Research8Hard multi-constraint discovery questions.

The scored split is private and rotates per release; the freshness generator is private, so providers cannot train on the test. A task passes only if it addresses the question, is correct, is current where that applies, and is supported by a returned source. Fabricated support scores minus 1: fabricated confidence is worse than an honest miss. Provider errors and timeouts count against the provider; harness faults on our side are voided and re-run, never scored.

Verdict Buckets

VerdictMeaning
Correct AnswerAnswered the task correctly.
Question Not AddressedNever addressed the question asked.
Incorrect AnswerAnswered confidently, but incorrectly.
Outdated AnswerAnswered with out-of-date information.
Fabricated CitationCited a source that does not support the claim.
Provider API ErrorThe provider call itself failed.

Baselines

Claude native WebSearch baseline
69.1
reference marker on every chart
Cost per 1k successful answers
$30.19
fees only: searches_made x $10 per 1,000 searches (Anthropic list price; one billed use per search regardless of result count, failed searches unbilled); payload ingestion is not separately measurable in this lane

A no-search baseline (the identical pinned agent with search disabled) scored 16.7 on the same bank.

Cost Method

Cost per 1,000 successful answers = billed provider fees across the whole workload (all calls, including retries and failed tasks) plus payload ingestion priced at the pinned shopper's input rate (Claude Opus 4.8, $5.00 per million tokens), divided by successful tasks.

Worked example (Firecrawl): $0.01 list price/call x 2.4 avg searches per task / 0.83 pass rate = $0.0379 per successful answer ($37.94 per 1,000).

Versioning

  • Kind weights and the judge (model and prompt) are pinned per index version; any change ships as a new, recalibrated version.
  • Task splits are private and rotate per release; saturated sources are retired with published rationale.
  • Changes and corrections are logged in the changelog, never silent.

Telemetry Semantics

  • Liveness is not "returns 200": every watched surface carries an expected status set, and a 200 with the wrong content records as a soft-404, not presence.
  • Blocked is not down: bot walls are recorded as blocked and never counted as downtime.
  • Percentiles come from rollups only; a single-sample latency is never displayed. Speed and liveness are measured with unpaid probes from a single vantage point (our probe region). Treat them as our view, not global ground truth.
  • The watcher never pays and never authenticates: unpaid, unauthenticated requests only.

Lineage

The judge design adapts the Universal Verifier's published calibration principles (Rosset et al., arXiv:2604.06240); hurdle and signed grounding follow Mercor's ACE. Task-bank seeds: SimpleQA-Verified (MIT), FRAMES and WebWalkerQA (Apache-2.0).