Past editions
Research archive
Every published edition, by research day and micro-vertical. Original SLED-wide editions keep their dates, issue numbers and analysis; later role enrichment is labeled separately.
Edition search matches summaries and pattern text. For a specific study, organization or source URL, search the evidence instead.
87 editions
All published research, newest first
September 13, 202610 editions
- Campus OperationsIssue 083 resources · 1 patternThree new archive additions examine IT chatbot benefit claims, administrative pilot expiry and independent tertiary IT control findings. One cross-source interpretation connects time-limited access to tested revocation. These undated or historical sources…
- A pilot end date should trigger a tested access decision
- College AthleticsIssue 082 resources · 1 patternTwo newly archived sources examine collegiate tennis injury classification and independent soccer-model scrutiny. One pattern focuses on error-aware acceptance tests. Neither demonstrates injury reduction or a release since the last completed run. Recruiting,…
- Test injury-class errors before accepting headline accuracy
- Emergency ServicesIssue 083 resources · 2 patternsThree newly catalogued historical sources examine a voice-agent stroke-assessment simulation, wildfire alert-authorization simulation and international EMS evidence gaps. Two patterns address decision validation and the difference between authorization and…
- Validate the local decision rather than inherit a published model score
- A review or authorization step must itself be evaluated
- K–12Issue 083 resources · 1 patternThree newly archived sources examine California district approval rules, teacher rubric-authoring friction and Hong Kong school AI planning. The workshop reports perceptions rather than retained learning or verified savings; official guidance describes…
- Make human review feasible within the actual workflow
- Local GovernmentIssue 083 resources · 1 patternThree new-to-archive sources cover Cumberland's reported process-automation gains, San José's curb-sensing validation gaps and NSW council AI inventory weaknesses. One pattern supports funding operational ownership before expansion. Recent case-study…
- Fund operational ownership before expanding automation and AI
- NVIDIAIssue 084 resources · 2 patternsFour newly covered sources examine NVIDIA's regional university hub participation, NSF's infrastructure funding boundary, September cloud operational requirements and independent H100 energy measurements. Two patterns connect partnership planning to funded…
- Fund the operating service behind the partnership
- Use reproducible measurements for service and energy decisions
- Public SafetyIssue 083 resources · 1 patternThree historical sources new to the archive cover police adoption perceptions, New York court pilots and Catalan corrections explainability. Self-reported benefits, operator accounts and academic governance proposals remain distinct from measured…
- Positive pilot feedback leaves the operational value question open
- ResearchIssue 083 resources · 1 patternThree newly covered sources examine research-infrastructure financing, shared-platform operational handoffs and scientific forecasting limits. July and May documents and an undated operator page are explicitly contextual, not same-day announcements. Program…
- Shared infrastructure requires explicit local and consortium responsibilities
- State GovernmentIssue 082 resources · 1 patternTwo newly archived historical sources examine North Carolina's department-wide AI rollout and academic scrutiny of vendor disclosures. Neither establishes department-wide gains or safer deployed systems. One pattern proposes workflow-specific acceptance…
- Carry unresolved evidence questions into each expanded workflow
- Student SuccessIssue 082 resources · 0 patternsTwo newly archived sources cover a UK staff–student perceptions study and U.S. advising controls. Neither demonstrates learning, retention or savings. The September 10 paper analyzes older survey data; Virginia Tech guidance is explicit February backfill. No…
- Campus OperationsIssue 083 resources · 1 patternThree new archive additions examine IT chatbot benefit claims, administrative pilot expiry and independent tertiary IT control findings. One cross-source interpretation connects time-limited access to tested revocation. These undated or historical sources…
September 12, 202610 editions
- Campus OperationsIssue 073 resources · 1 patternThree newly archived sources examine embedded administrative AI readiness, IT/HR service responsibility and staff adoption denominators. One cross-source interpretation distinguishes feature availability, activation and actual use. These older sources fill…
- An adoption count needs an explicit stage and denominator
- College AthleticsIssue 072 resources · 0 patternsTwo newly archived sources cover Morgan's funded student-athlete chatbot plan and international scrutiny of soccer recruiting models. Neither establishes measured collegiate benefit or a new-since-last-run release. Zero cross-source patterns; independent…
- Emergency ServicesIssue 073 resources · 2 patternsThree newly catalogued historical sources cover New York EMS AI-use documentation, industrial fire digital-twin simulation and a multi-hazard AI review with reporting limitations. Two patterns address authorization boundaries and validation scope. Role…
- Make the transition from AI assistance to authorized action explicit
- Keep demonstration scope visible when assessing an integrated disaster platform
- K–12Issue 072 resources · 1 patternTwo newly archived research reviews examine independent learning and teachers' capacity to act on AI outputs. Mixed results and methodological limits support bounded local evaluation, not blanket adoption claims. One cross-source pattern connects learning and…
- Evaluate learner independence alongside the capacity to provide support
- Local GovernmentIssue 073 resources · 1 patternThree new-to-archive historical sources cover Wigan's reported administrative gains, Atlanta's resident-facing 311 limitations and Maricopa's qualified AI audit findings. One pattern supports checking whether processed information leads to useful service…
- Check whether processed information leads to useful service action
- NVIDIAIssue 073 resources · 2 patternsThree newly covered sources examine mixed RAG reasoning results, independent evaluation-method limits and September NIM support notices. Two patterns connect task-level quality assurance to configuration and migration decisions. Vendor benchmark scores, an…
- Calibrate answer-quality decisions against the intended task
- Carry answer-quality tests into support-driven migrations
- Public SafetyIssue 073 resources · 0 patternsThree new-to-archive sources cover September 11 reporting on fabricated testimony in a New Mexico appeal, a historical Virginia Beach body-camera AI replication study, and September 10 UK police-registry commitments. Findings distinguish a documented failure,…
- ResearchIssue 073 resources · 1 patternThree newly covered sources examine research-agent execution failures, a qualified workshop-review milestone and planned safeguards for scientific instruments. A September 10 benchmark is paired with clearly dated March and July context. Rubric scores,…
- Define acceptance around scientific evidence, not the appearance of completed work
- State GovernmentIssue 072 resources · 1 patternTwo newly archived sources examine Mississippi's synthetic certificate-agent demonstration and NSW's AI registration and assessment transition. Operator results do not establish production reliability; mandatory guidance does not prove compliance or benefit.…
- Carry prototype boundaries into the operating risk record
- Student SuccessIssue 073 resources · 0 patternsThree new archive sources cover a U.S. course-outreach RCT, German assistant usage and conceptual limits on educational-agent authority. Grade-threshold gains have weaker adjusted evidence; adoption is not learning and proposed safeguards remain unvalidated.…
- Campus OperationsIssue 073 resources · 1 patternThree newly archived sources examine embedded administrative AI readiness, IT/HR service responsibility and staff adoption denominators. One cross-source interpretation distinguishes feature availability, activation and actual use. These older sources fill…
September 11, 202610 editions
- Campus OperationsIssue 063 resources · 1 patternThree new archive additions examine simulated data-center cooling gains, La Trobe's campus energy measurement platform and a preprint review of validation gaps. One cross-source interpretation separates subsystem optimization from a complete campus investment…
- A cooling objective is only part of a campus investment result
- College AthleticsIssue 063 resources · 1 patternThree newly archived sources cover BYU's sports-video development workflow, Towson's announced staff-contract AI access, and international sprint-screening limits. One supported pattern calls for checking headline descriptions against implemented methods. No…
- Check implementation details against headline model descriptions
- Emergency ServicesIssue 063 resources · 2 patternsThree newly catalogued sources cover FDNY ambulance-planning simulation, NASEMSO documentation controls and Nashville's Sim911 training deployment. Two patterns distinguish simulation evidence from operational outcomes and make corrections auditable.…
- Evaluate transfer from a useful simulation to operational performance
- Include human correction in the workflow and cost model
- K–12Issue 063 resources · 1 patternThree newly archived sources connect teacher-supported writing feedback, current middle-school AI literacy expansion and international child-specific chatbot safeguards. Partnership findings and operator reach do not establish durable learning; UNICEF…
- Plan educator support as part of the implementation
- Local GovernmentIssue 063 resources · 1 patternThree new-to-archive sources cover a municipal budget-assistant prototype, utility decision-reasoning scrutiny and recent Weld County AI-infrastructure oversight. One pattern supports testing evidence validity in local workflows. Historical prototype results…
- Validate the supporting evidence within the local task
- NVIDIAIssue 064 resources · 1 patternFour newly covered sources examine speculative-decoding performance, evaluation weaknesses, TensorRT-LLM configuration limits and September 9 driver upgrade dependencies. One pattern connects serving acceptance to actual workload and load. Academic funding,…
- Approve speculative decoding against the intended workload and load
- Public SafetyIssue 063 resources · 1 patternThree historical sources new to the archive cover the UK PoliceAI pilot programme, a U.S. prosecution-memo experiment, and a COMPAS dataset reanalysis. Operator projections remain distinct from measured results; old-model and single-county limitations are…
- Test multiple dimensions of bias rather than relying on one reassuring result
- ResearchIssue 063 resources · 0 patternsThree newly covered sources distinguish announced public-university research capacity, uneven scientific-agent benchmark results, and qualitative guidance for maintaining trustworthy research services. The September 10 Morgan State announcement is paired with…
- State GovernmentIssue 062 resources · 1 patternTwo newly archived historical sources examine Missouri transport ML's operational and economic limits and Australian AI disclosure quality. Technical scores and published assurances do not establish service fitness. Role guidance and one cross-source…
- Connect assurance evidence to the operational decision
- Student SuccessIssue 063 resources · 1 patternThree new archive sources cover a U.S. course-chatbot null trial, selective student AI use in Australia and Irish teaching-and-learning controls. None establishes durable learning, retention or savings. The September 9 survey is recent evidence; the trial and…
- Evaluate learner memory through educational and data-control reviews
- Campus OperationsIssue 063 resources · 1 patternThree new archive additions examine simulated data-center cooling gains, La Trobe's campus energy measurement platform and a preprint review of validation gaps. One cross-source interpretation separates subsystem optimization from a complete campus investment…
September 10, 202610 editions
- Campus OperationsIssue 053 resources · 1 patternThree newly archived sources examine campus benefit measurement and governance: UT Austin's salary-valued staff capacity, UAEU's analytically modeled HR workflow gains, and an international review of governance-outcome evidence. One cross-source…
- State what was measured before putting a dollar value on a benefit
- College AthleticsIssue 053 resources · 1 patternThree newly archived sources cover Division II qualifier-processing errors and correction, Toledo's reported staff training and coaching workflows, and Australian sport-specific operating controls. One supported pattern emphasizes testing exception and appeal…
- Test exception and appeal paths before relying on routine success
- Emergency ServicesIssue 056 resources · 3 patternsSix newly catalogued sources cover September 10 Seattle EMS oversight reporting, California satellite commissioning and AI camera claims, NASA airborne wildfire research, an official DHS audit summary, international susceptibility-validation limits and RAND's…
- Measure whether assistance reaches a completed service handoff
- Match reported metrics to the decision and test population
- Treat service availability and organizational readiness as separate gates
- K–12Issue 054 resources · 2 patternsFour newly archived sources examine assessment reliability, tutoring workflow trade-offs, LAUSD student-access restrictions and school staff data handling. Original research distinguishes scoring accuracy from agreement; an operator paper reports mixed…
- Validate the AI measure before using it to judge the intervention
- Define student access and staff data handling as separate operating decisions
- Local GovernmentIssue 053 resources · 2 patternsThree new-to-archive sources connect San Francisco's historical AI rollout and unfinished benefit-measurement commitments, Portland's 2026 controls for embedded and internally developed AI, and independent research on procurement-checklist limits. Two…
- Give review owners the expertise and capacity to act
- Keep AI review connected to changes after purchase
- NVIDIAIssue 054 resources · 1 patternFour sources cover September 10 NIM serving results and the Palantir supply-chain collaboration, alongside newly relevant cache-security guidance and independent research. One pattern links capacity acceptance to intended tenant isolation. Vendor claims,…
- Measure capacity under the intended information-sharing policy
- Public SafetyIssue 054 resources · 1 patternFour new-to-archive sources cover Bridgewater's surveillance-contract cancellation, EgoPolice video-understanding limits, a historical prison–court handoff review, and Western Australia's facial-recognition trial controls. U.S. evidence and transferable…
- Authorization should be revisited throughout a surveillance system's life
- ResearchIssue 053 resources · 0 patternsThree newly archived sources cover CU Boulder's deadline-relevant AI research mentoring pilot, an academic documentation-cost study and NSF grant-review controls. Program intentions, proxy measurements and policy requirements remain distinct. Older and…
- State GovernmentIssue 053 resources · 2 patternsThree newly archived sources examine California's shared AI platform, North Carolina's forthcoming oversight playbook and independent scrutiny of state AI contract provisions. Platform capability claims and planned controls are not measured service gains.…
- Shared model access needs evidence for each material change
- Approval needs an operating review capacity
- Student SuccessIssue 052 resources · 0 patternsTwo newly archived sources examine programming-tutor feedback and learner judgment. A Berkeley deployment comparison finds mixed correct application despite stronger feedback uptake; a recent critical review provides unvalidated diagnostic questions, not…
- Campus OperationsIssue 053 resources · 1 patternThree newly archived sources examine campus benefit measurement and governance: UT Austin's salary-valued staff capacity, UAEU's analytically modeled HR workflow gains, and an international review of governance-outcome evidence. One cross-source…
September 9, 202610 editions
- Campus OperationsIssue 043 resources · 1 patternThree newly archived sources cover Buffalo's planned AI-center heat recovery, NC State's facilities analytics service and Jisc's agent-autonomy discussion. None demonstrates causal AI savings. One cross-source interpretation addresses the transition from…
- The transition from recommendation to action needs a named owner
- College AthleticsIssue 042 resources · 1 patternTwo newly archived sources distinguish Rice's operational dashboard from predictive development and test the limits of broadcast AI tracking. One supported pattern calls for separate validation of measurements and downstream predictions. No new-since-last-run…
- Validate measurement and prediction as separate layers
- Emergency ServicesIssue 044 resources · 2 patternsFour newly catalogued sources examine simulated ambulance deployment, ECMWF cyclone intensity correction, NHC operational integration and a controlled extreme-storm generalization test. Evidence supports bounded evaluation and expert oversight, not guaranteed…
- Average forecast improvement does not settle extreme-event reliability
- Define the handoff from experimental guidance to authorized action
- K–12Issue 043 resources · 2 patternsThree newly archived sources examine the scope of a September 9 school AI privacy announcement, Kentucky's older implementation inventory and parent participation in AI decisions. The original agreement qualifies broad promotional claims; implementation lists…
- Match each actual workflow to its evidence and protections
- Connect family participation to a decision that can still change
- Local GovernmentIssue 043 resources · 2 patternsThree new-to-archive sources connect municipal zoning-answer quality, a funded residential-permitting test plan, and new UK grid-AI guidance relevant to municipal utilities. Two patterns support useful-outcome testing and explicit authority boundaries. No…
- Test useful resident outcomes before public launch
- Set authority by demonstrated performance and recoverability
- NVIDIAIssue 044 resources · 2 patternsFour newly covered sources examine NVIDIA Retriever placement and scheduling, NeMo template governance, independent template-backdoor research and a September 9 Australian DSX expansion announcement. Two patterns distinguish data location from behavioral…
- Data location and instruction integrity require separate controls
- Infrastructure scale does not establish workload readiness
- Public SafetyIssue 043 resources · 1 patternThree new-to-archive sources cover court and probation implementation plans, U.S. supervision evidence review, and recent police workforce-monitoring scrutiny. Historical announcements and legal analysis are dated explicitly; no fresh causal benefit or…
- Automated assertions need a usable route for contextual review
- ResearchIssue 043 resources · 0 patternsThree newly archived sources cover planned U.S. university research capacity, a peer-reviewed clinical-data agent evaluation and an exploratory scientific-discovery benchmark. Benefits, failure modes and evaluator limitations remain distinct. September 8 and…
- State GovernmentIssue 043 resources · 2 patternsThree new archive records connect California's September 9 AskCA announcement with historical North Carolina workforce evaluation and New York governance audit evidence. AskCA is recruiting testers, not reporting demonstrated service improvements. NC savings…
- Measure completed work and verification effort before scaling
- Assign evidence ownership alongside product ownership
- Student SuccessIssue 043 resources · 0 patternsThree newly archived sources cover mentored AI capstones, delayed independent learning and student access. UCF's September 9 implementation account is promotional and has no measured educational outcomes; a 2025 retention study reports task-specific mixed…
- Campus OperationsIssue 043 resources · 1 patternThree newly archived sources cover Buffalo's planned AI-center heat recovery, NC State's facilities analytics service and Jisc's agent-autonomy discussion. None demonstrates causal AI savings. One cross-source interpretation addresses the transition from…