We analyzed every issue of AI Weekly — 476 issues, 760,000+ words, 11 years of human-curated AI coverage — and cross-referenced it with engagement data from 60,000+ readers across two newsletter platforms. The result is the most comprehensive longitudinal view of the AI industry available anywhere.
Below you'll find what the industry talked about, when it stopped, which companies dominated, what readers actually cared about, and how the emotional tone of AI coverage shifted from optimism to fear to money. Every data point is sourced from real editorial decisions and real reader behavior — not surveys, not funding databases, not keyword scrapers.
This Week's Signals
The Pulse
The AI Momentum Index is a composite of three forces: innovation (new models, architectures, research breakthroughs), investment (funding rounds, valuations, startup activity), and disruption (layoffs, regulation, deepfakes). When all three rise together — as they did in 2023 — the industry is in a gold rush. When disruption rises faster than innovation — as in Q1 2026 — the consequences of AI are becoming as prominent as the technology itself.
The emotional register of AI coverage has fundamentally changed. In 2015, practical language dominated — "how to," "deploy," "use case" appeared 4.7 times per issue. By 2026, fear language ("danger," "threat," "risk") has exploded 12x from 0.3 to 3.6 per issue. Money language ("billion," "valuation," "funding") tripled. Geopolitics barely existed before 2022 — now it's 2.2 mentions per issue. AI stopped being a technology story and became a power story.
Concept velocity measures what's accelerating and what's fading, comparing 2025-2026 coverage against 2023-2024 baselines. "Agents" surged 1,003% — the fastest-rising concept in our dataset. "World models" (+788%) reflects Yann LeCun's AMI Labs and the post-LLM movement. "Hallucination" (+612%) went from ignored to critical concern. On the other side: diffusion models, RAG, and transformers have all disappeared from discourse — not because they failed, but because they became invisible infrastructure. The terms that vanish fastest are often the ones that won.
The Company Race
Google dominated AI coverage for 7 consecutive years (2015-2021), peaking at 4.25 mentions per issue in 2017. OpenAI overtook them in 2023 and has led since. But the real story is Q1 2026: Anthropic surpassed OpenAI for the first time (1.72 vs 1.22 mentions per issue), driven by the #QuitGPT movement and Claude hitting #1 on the App Store. NVIDIA's rise from near-zero to 1.0 per issue mirrors the infrastructure investment wave. DeepMind has been in steady decline since the Google Brain merger.
Issue titles are what subscribers see in their inbox — they drive the open decision. What makes the subject line reveals what the editor thinks will get attention. Robotics leads with 24 title appearances across 11 years — it's consistently our most headline-worthy topic. Chatbots dominated titles in 2023 (the ChatGPT year) but have already faded. Ethics appeared in 11 headlines. The shift from technology titles (robotics, chatbots) to consequence titles (regulation, funding, jobs) mirrors the industry's maturation from lab to boardroom.
Concepts & Hype Cycles
"Agent" appeared 0.2 times per issue in 2023. In Q1 2026, it's 3.83 — a 1,003% increase. No other term in our 11-year dataset has risen this fast. For context: "ChatGPT" peaked at 2.58 per issue in 2023.
"Generative AI" peaked at 1.67 per issue in 2023 and dropped to 0.17 in Q1 2026 — a 90% decline. A textbook hype cycle captured in real time. "LLM" peaked later (2024) and is declining more slowly, suggesting it has more staying power as a technical term.
The crossover happened in 2025. Hype language (generative AI + ChatGPT + GPT) has been declining since its 2023 peak. Business reality language (revenue + valuations + funding) has been rising since 2024. In Q1 2026, for the first time, the industry talks more about money than about magic. The "show me the money" era has arrived.
The full concept map reveals distinct eras. 2015-2018: the Autonomous Vehicle era (China, funding, self-driving dominated discourse). 2019-2022: the Robotics & Ethics era (the two most consistent topics). 2023: the Chatbot spike (a singularity of attention that distorted everything else). 2024-2025: the Regulation build-up. 2026: the Agent revolution, with safety and military AI rising alongside it. Each era lasted 2-3 years. We are currently 6 months into the Agent era.
Concerns & Society
Ethics dominated the concern landscape from 2019 to 2022, peaking at 3.2 mentions per issue (the GPT-2 "too dangerous to release" debate, Timnit Gebru's firing, EU AI Act drafting). But ethics coverage has fallen 55% since its peak — replaced by "safety" and "regulation" as the discourse shifted from philosophical questions to practical governance. Deepfakes spiked in Q1 2026 (0.78/issue, up from 0.09 in 2023) as AI-generated political ads entered the US midterms. Our engagement data shows that ethics-focused content has a 17.6% open rate with a 20.8% click-through — readers who care about AI ethics are among the most engaged in our entire audience.
Technology Waves
Each technology wave follows the same pattern: emergence, peak attention, absorption into baseline, disappearance from discourse. "Machine learning" peaked in 2021 (3.72/issue), now 0 in 2026 — not because it stopped mattering, but because it became assumed. "Deep learning" followed the same curve 2 years later. "Transformers" peaked in 2022 (0.44/issue) and are now at 0. "Diffusion models" lasted just 2 years (2022-2023). The current wave — agents, reasoning, world models — is in the emergence-to-peak phase. History says it has 18-24 months before it, too, becomes invisible infrastructure.
Audience & Engagement
From 30 subscribers in May 2015 to 37,000 in 2025. The fastest growth came in 2017 (+129%), when AI transitioned from research curiosity to front-page news. Growth plateaued in 2021-2022 (the "AI winter of attention" between GPT-3 and ChatGPT). The 2023 resurgence (+39%) was entirely driven by ChatGPT — proving that a single product can reshape an entire field's public profile.
General AI (20,735 subscribers) dominates. But size doesn't equal engagement. AI for Good, with just 2,745 subscribers, has a 45.3% click-through rate — the most engaged niche in our entire dataset. Ethics subscribers (7,448) click through at 20.8%. The pattern is clear: the smallest audiences are the most passionate. Broad topics attract casual readers. Specific topics attract committed ones.
When subscriber growth is plotted against publication frequency, there's no correlation — years with more issues didn't grow faster. What does correlate: content relevance. The three fastest-growth years (2016: +86%, 2017: +129%, 2023: +39%) each coincided with a moment where AI became newly relevant to non-specialists. In 2016-2017, it was autonomous vehicles and the China AI race. In 2023, it was ChatGPT. Growth doesn't come from publishing more. It comes from the world caring more.
AI Across Industries
Robotics has been the most consistently covered industry application since 2019, peaking at 4.15 mentions/issue in 2024 — and it's our third-largest subscriber topic (9,163). Healthcare peaked in 2024 (1.6/issue) as the initial wave of medical AI hype normalized. The biggest 2026 shifts: finance surged to 0.78/issue (7x its 2024 level), and military AI rose to 0.39/issue as AI moved from labs to battlefields. Legal AI is accelerating as AI-generated evidence enters courtrooms.
The quarterly view reveals the speed of change within the AI industry. Agents barely registered in early 2023 but dominate every quarter of 2026. Regulation spiked in Q3-Q4 2023 (EU AI Act passage) then stabilized. Deepfakes surge in Q1 2026 as synthetic political media enters the US midterms. Safety discourse tripled between Q4 2024 and Q1 2026 as frontier labs faced increasing scrutiny. The quarterly cadence shows how fast consensus forms — and how fast it shifts.
Methodology
Content analysis. Term frequency across all 476 issues of AI Weekly (May 2015 – March 2026), totaling 760,000+ words. We track 20 companies, 30 concepts, 8 concern categories, 7 industries, and 5 emotional themes. Mentions in issue titles are weighted 5x (titles drive the open decision and reflect editorial judgment about significance). Body text mentions are weighted 1x. All figures are normalized per issue per year to account for varying publication frequency.
Concept velocity. Compares average per-issue mention rates in 2025-2026 against 2023-2024 baselines. A velocity of +100% means the concept is mentioned twice as often per issue as it was two years ago.
Theme analysis. Categorizes language into five emotional registers — hype, fear, money, geopolitics, and practical — using curated keyword dictionaries. Tracks how the tone of AI coverage evolved over a decade.
Engagement data. Derived from 401,000+ email delivery events across 20,000+ unique recipients and a combined audience of 61,760 across two newsletter platforms. Open rates, click-through rates, and topic engagement are measured at the subject-line level.
Subscriber data. 511 weekly data points covering growth from 30 to 37,000 subscribers. Topic interest reflects active subscriber preferences as of Q1 2026.
Composite index. The AI Momentum Index is a weighted sum: Innovation (model releases + architecture mentions × 2), Investment (funding + valuations + startup activity × 3), Disruption (layoffs + regulation + deepfakes × 2). Weights reflect relative editorial significance.
This page updates automatically with each new issue. The full dataset will be available for download in Q2 2026.