What to follow next

A short list of people, labs, reports, and books that show where cheap intelligence is going. It stays deliberately short: 25 entries instead of an encyclopedia. Verified: 2026-08-18. Review once a quarter because the field changes. Courses and channels have a separate page under Recommended resources.

People worth reading regularly

Ethan Mollick: the One Useful Thing newsletter covers work and AI without hype. His view is close to this book.
Andrew Ng: The Batch brings an engineer's view of what works in practice.
Andrej Karpathy: lectures and essays explain models and "software 3.0" in plain language.
Dario Amodei of Anthropic and Demis Hassabis of Google DeepMind: essays and interviews give a first-hand view of capability boundaries. Read them as positions from their labs, not neutral forecasts.
Mustafa Suleyman of Microsoft AI: social effects and containment.
Fei-Fei Li: spatial and physical AI.

Labs and companies to follow for releases and data

Anthropic, including the Economic Index, publishes data on how people use AI at work.
OpenAI, including the GDPval benchmark used in our profession analysis.
Google DeepMind and Meta AI for research at the frontier.
NVIDIA and Microsoft for the rails. Their reports show who earns infrastructure rent. See the rails map.

Reports to read once a year

Stanford AI Index: an annual data review of the field.
WEF Future of Jobs: employer surveys on skills that are rising or falling.
McKinsey State of AI: what companies implement and where they get stuck.
Anthropic Economic Index: tasks that people actually delegate to AI.

Books for further reading

Ethan Mollick, Co-Intelligence: working with AI as a thinking partner.
Mustafa Suleyman, The Coming Wave: the technology wave and attempts to contain it.
Reid Hoffman, Superagency: an optimistic frame for AI as a way to expand human agency.

Places that measure, rather than announce

Everything above is people and companies. This shorter list is instruments: primary series you can check yourself, without anyone's press release in between. They settled several arguments during our own August 2026 review.

Cloudflare Radar for what share of web traffic is machines, and what the crawlers came for.
ARC Prize for whether models can handle genuinely unfamiliar problems, as opposed to familiar kinds of work. Its current benchmark is the sharpest public counterweight to capability marketing.
Eurostat and your own national statistics office for labour-market claims. Circulating figures about collapsing graduate hiring rarely survive contact with the official series.
European Commission digital strategy pages for what AI regulation actually requires today, now that the EU AI Act's transparency duties are being enforced.

The habit matters more than the list: when a number surprises you, find the instrument that produced it before repeating it.

Do not subscribe to everything. Choose one person, one report, and one book each quarter. That is enough to follow the field. The rest is information noise.

Back: Live pages index · Recommended resources. Version: 2026-08-18.