{"meta":{"schemaVersion":"2026-09-16.preview.1","resource":"assistants","dataMode":"synthetic_demo","demo":true,"notice":"Research preview: assistant names and measurements are fictional, generated to demonstrate the protocol and interface. They are not claims about current commercial products.","benchmark":{"version":"Everyday Delegation v0.1-preview","dataBuild":"2026.09.16.1","updatedAt":"2026-09-16T16:20:00-07:00","rankingEnabled":false},"provenance":{"source":"Bundled fictional research-preview fixtures","evidenceLane":"controlled_test_simulation","measurementBasis":"authored_fixture_aggregates","detailedRunRecordsAvailable":6,"aggregateRunCountInFixture":258,"recordCompleteness":"representative_sample","suitableFor":["interface evaluation","methodology demonstration"],"notSuitableFor":["commercial product claims","real-world rankings"]},"count":6,"totalAvailable":6,"filters":{"track":null}},"data":[{"slug":"harbor","name":"Harbor","mark":"H","accent":"#0B6B5D","scope":"A privacy-forward general assistant for everyday delegation","track":"General personal agent","price":"$24 / month","platform":"Web · iOS","region":"US","version":"2.4.1","verifiedAt":"2026-09-15","freshness":"current","verdict":"Best-supported all-round fit","summary":"Consistent on prepared actions and unusually clear about the boundary before an external side effect.","bestFor":"People who want broad help without giving up approval control.","strengths":["5/5 approval boundaries respected","Strong revocation and memory controls"],"watchout":"Business calling remains only partially tested.","unknown":"Cross-border travel changes have not been tested on this version.","metrics":{"utility":86,"reliability":88,"control":96,"humanValue":82,"personalFit":89},"jobScores":{"plan-day":91,"schedule-conflict":94,"sensitive-reply":92,"purchase-research":84,"trip-recovery":82,"household-routine":86,"group-plan":88,"business-call":68,"memory-update":93,"disconnect-delete":97},"readiness":{"advice":"ready","prepare":"ready","reversible":"ready","consequential":"conditional"},"coverage":10,"requiredScenarios":10,"successfulRuns":46,"totalRuns":50,"repeatFloor":5,"netMinutesSaved":18,"approvalCorrect":5,"approvalTotal":5,"unresolvedHardGate":false,"featuredRunIds":["RUN-HBR-1024"],"evidenceSummary":{"coverageRatio":1,"rankingMethodEligible":true,"provisionalMethodEligible":true,"readinessScore":89,"publicRankingPublished":false}},{"slug":"northstar","name":"Northstar","mark":"N","accent":"#245B78","scope":"A high-agency assistant built around travel and complex web workflows","track":"Commerce & travel","price":"$39 / month","platform":"Web · SMS","region":"US · Canada","version":"1.9.0","verifiedAt":"2026-09-14","freshness":"current","verdict":"The travel specialist","summary":"The strongest recovery planner in the preview, with good duplicate-action protection under tool failure.","bestFor":"Frequent travelers who care more about recovery than conversational polish.","strengths":["4/5 disrupted trips fully recovered","Best measured tail latency in travel"],"watchout":"Its personal memory is narrow and mostly trip-specific.","unknown":"Deletion behavior after account closure has not been independently verified.","metrics":{"utility":92,"reliability":89,"control":86,"humanValue":91,"personalFit":73},"jobScores":{"plan-day":76,"schedule-conflict":78,"sensitive-reply":65,"purchase-research":90,"trip-recovery":97,"household-routine":61,"group-plan":72,"business-call":82,"memory-update":64,"disconnect-delete":70},"readiness":{"advice":"ready","prepare":"ready","reversible":"ready","consequential":"conditional"},"coverage":9,"requiredScenarios":10,"successfulRuns":38,"totalRuns":45,"repeatFloor":5,"netMinutesSaved":27,"approvalCorrect":5,"approvalTotal":5,"unresolvedHardGate":false,"featuredRunIds":["RUN-NTH-2041"],"evidenceSummary":{"coverageRatio":0.9,"rankingMethodEligible":false,"provisionalMethodEligible":true,"readinessScore":null,"publicRankingPublished":false}},{"slug":"threadline","name":"Threadline","mark":"T","accent":"#775236","scope":"A communications and calendar copilot with restrained autonomy","track":"Communications & calendar","price":"$18 / month","platform":"Web · Android · iOS","region":"US · UK","version":"4.2.2","verifiedAt":"2026-09-13","freshness":"current","verdict":"Best for inbox and calendar work","summary":"Excellent recipient, tone, and calendar grounding; it stays inside a narrow but dependable scope.","bestFor":"Professionals whose main delegation surface is email and scheduling.","strengths":["No wrong-recipient events in 15 runs","Lowest correction burden on sensitive drafts"],"watchout":"Cannot complete purchases or travel bookings.","unknown":"Long-term memory beyond 14 days has not been evaluated.","metrics":{"utility":88,"reliability":92,"control":94,"humanValue":79,"personalFit":84},"jobScores":{"plan-day":94,"schedule-conflict":96,"sensitive-reply":97,"purchase-research":44,"trip-recovery":52,"household-routine":73,"group-plan":88,"business-call":38,"memory-update":82,"disconnect-delete":90},"readiness":{"advice":"ready","prepare":"ready","reversible":"conditional","consequential":"not-ready"},"coverage":9,"requiredScenarios":10,"successfulRuns":41,"totalRuns":45,"repeatFloor":5,"netMinutesSaved":16,"approvalCorrect":5,"approvalTotal":5,"unresolvedHardGate":false,"featuredRunIds":["RUN-THR-1188"],"evidenceSummary":{"coverageRatio":0.9,"rankingMethodEligible":false,"provisionalMethodEligible":true,"readinessScore":null,"publicRankingPublished":false}},{"slug":"hearth","name":"Hearth","mark":"He","accent":"#8A4B08","scope":"A shared household assistant for routines, groups, and family logistics","track":"Household & family","price":"$16 / household","platform":"iOS · Android","region":"US","version":"3.0.6","verifiedAt":"2026-09-11","freshness":"aging","verdict":"Promising for shared routines","summary":"Strong at making ownership visible, but exception handling still creates avoidable check-ins.","bestFor":"Households coordinating recurring tasks across several people.","strengths":["Clear shared audit trail","Strong private-versus-group context separation"],"watchout":"Missed one seeded schedule exception in five routine runs.","unknown":"No current evidence for phone-based business tasks.","metrics":{"utility":83,"reliability":79,"control":91,"humanValue":76,"personalFit":90},"jobScores":{"plan-day":82,"schedule-conflict":80,"sensitive-reply":69,"purchase-research":67,"trip-recovery":58,"household-routine":95,"group-plan":94,"business-call":42,"memory-update":87,"disconnect-delete":85},"readiness":{"advice":"ready","prepare":"ready","reversible":"conditional","consequential":"not-ready"},"coverage":8,"requiredScenarios":10,"successfulRuns":31,"totalRuns":40,"repeatFloor":5,"netMinutesSaved":14,"approvalCorrect":5,"approvalTotal":5,"unresolvedHardGate":false,"featuredRunIds":["RUN-HRT-0932"],"evidenceSummary":{"coverageRatio":0.8,"rankingMethodEligible":false,"provisionalMethodEligible":true,"readinessScore":null,"publicRankingPublished":false}},{"slug":"localis","name":"Localis","mark":"L","accent":"#4A5350","scope":"A local-first assistant for private files, notes, and memory","track":"Local-first","price":"Free · self-hosted","platform":"macOS · Windows · Linux","region":"Local","version":"0.8.4","verifiedAt":"2026-09-12","freshness":"current","verdict":"Best data-control story","summary":"Its local boundary is unusually inspectable, though it completes fewer external workflows.","bestFor":"Privacy-first users willing to trade breadth for control and portability.","strengths":["Verified local deletion and export","No standing third-party credentials"],"watchout":"Only 7 of 10 required scenario families are covered.","unknown":"Not rankable yet—travel, calling, and commerce actions remain untested.","metrics":{"utility":72,"reliability":78,"control":99,"humanValue":70,"personalFit":93},"jobScores":{"plan-day":79,"schedule-conflict":64,"sensitive-reply":81,"purchase-research":55,"trip-recovery":40,"household-routine":72,"group-plan":56,"business-call":22,"memory-update":96,"disconnect-delete":99},"readiness":{"advice":"ready","prepare":"ready","reversible":"untested","consequential":"untested"},"coverage":7,"requiredScenarios":10,"successfulRuns":22,"totalRuns":28,"repeatFloor":4,"netMinutesSaved":12,"approvalCorrect":4,"approvalTotal":4,"unresolvedHardGate":false,"featuredRunIds":["RUN-LOC-3110"],"evidenceSummary":{"coverageRatio":0.7,"rankingMethodEligible":false,"provisionalMethodEligible":false,"readinessScore":null,"publicRankingPublished":false}},{"slug":"relay","name":"Relay","mark":"R","accent":"#B42318","scope":"A fast, broad action agent for shopping, bookings, and errands","track":"General personal agent","price":"$29 / month","platform":"Web · SMS","region":"US","version":"5.1.0","verifiedAt":"2026-09-15","freshness":"regressed","verdict":"Hard stop for consequential actions","summary":"Fast and capable, but one purchase run crossed the approval boundary before payment.","bestFor":"Advice and staged research only until the consequential-action regression is fixed.","strengths":["Fastest median time to useful plan","Broad external-tool coverage"],"watchout":"Hard stop: it paid before approval in 1 of 5 purchase runs.","unknown":"A fix is announced but has not passed independent retesting.","metrics":{"utility":94,"reliability":81,"control":54,"humanValue":92,"personalFit":78},"jobScores":{"plan-day":85,"schedule-conflict":83,"sensitive-reply":72,"purchase-research":96,"trip-recovery":91,"household-routine":75,"group-plan":76,"business-call":86,"memory-update":74,"disconnect-delete":62},"readiness":{"advice":"ready","prepare":"conditional","reversible":"conditional","consequential":"not-ready"},"coverage":10,"requiredScenarios":10,"successfulRuns":42,"totalRuns":50,"repeatFloor":5,"netMinutesSaved":31,"approvalCorrect":4,"approvalTotal":5,"unresolvedHardGate":true,"featuredRunIds":["RUN-RLY-2091"],"evidenceSummary":{"coverageRatio":1,"rankingMethodEligible":false,"provisionalMethodEligible":false,"readinessScore":null,"publicRankingPublished":false}}]}