Voices to follow
The AI news that matters is written by these people.
If you follow mainstream tech press for AI news, you will spend most of your time reading translations of what these people already wrote. The translation layer adds noise. This page is the source.
The list is curated the same way the book is written: with a specific thesis about what matters. Practitioners who actually operate agentic systems at the top. AGI-skeptic counter-voices next. Exec-translation writers last. No lab heads, no CEO Twitter, no "AI influencers" — those are available everywhere else.
This is the short version. The Further Reading page goes deeper, with primary works and context for each source.
Practitioners (the primary follow list)
Harness and operator voices. The people articulating the frame the book is built on. Most of these are on X and on Substack; a few write only on their personal sites. Subscribing to all of them takes about ten minutes.
Originator of the autoresearch / loop pattern the book is built on. Follow him to see the operator frame being invented in real time.
The most credible academic voice on AI-in-the-workplace for an executive audience. Coined the three-tier model / app / harness taxonomy.
The most prolific public-facing operator. His OS analogy (model=CPU, context=RAM, harness=OS) is the clearest teaching frame for executives.
Coined "AI Engineer." Translates practitioner work into industry concepts faster than anyone. Closest existing public voice to the book's "operator" identity.
Deepest writing on real LLM evals, error analysis, and the gap between demo and production. If your team is stuck in pilot purgatory, read him.
Applied-LLM patterns blog cited by every working ML team. Shipping patterns, LLM-as-judge, production traps.
Writes daily on agent / harness engineering. Useful for tracking the discipline as it evolves. Documented the vocabulary crystallizing in 2026.
Google DeepMind engineer. Practitioner-focused breakdowns of agent harness patterns. Clean articulations without vendor bias.
Historically names software-engineering disciplines more than he invents them. His 2026-04 treatment of harness engineering as a formal discipline is the mainstream-engineering legitimization.
Wrote the single cleanest expository piece on the harness-as-agentic-moat thesis (2026-03-27). If you read one piece on why the harness is the moat, read this one.
Box CEO. Named the "agent-manager role" publicly on 2026-04-15 — the first vendor-tier articulation that the builder-leader pattern has a hireable seat.
Vercel CEO. Named the "AI Software Factory" pattern on 2026-04-15 — the architectural shape of a Group 3 operator team producing continuous shippable output.
Five-architecture taxonomy of agent harnesses. Useful when the six components start to feel abstract and you want another structural frame.
Skeptics (the counter-case)
The book argues against the AGI-imminence frame, but it does so without strawmanning. These voices give the strongest, most defensible version of the skeptic case. Read them even if you disagree, and especially if you agree too quickly with the practitioners above.
Most prolific named AGI skeptic. His 2026-04-11 concession that agentic harness systems are "the single biggest advance in AI since the LLM" is a rare crossing-of-camps worth reading.
Raised $1.03B for AMI Labs on the "LLMs won't reach AGI" thesis. Career risk gives his skepticism credibility that executives cannot easily dismiss.
Rigorous on the gap between benchmark performance and real understanding. Santa Fe Institute. The strongest academic version of the skeptic case.
The strongest critique of LLM-as-cognition framing. Co-author of "On the Dangers of Stochastic Parrots."
Most rigorous AGI-definition voice. Arc Prize and ARC-AGI benchmark give empirical ballast for "current systems are not AGI" without becoming a hater.
Exec-translation writers
Less technical but closer to how the audience the book is for actually reads. Useful for sense-checking how the ideas land with executives who will not read the practitioners directly.
Every newsletter. The strongest signal on "what's landing with the executive audience." His onboarding-Claudie work is the cleanest exec-audience articulation of agent onboarding.
UK-based, executive-focused. Writes at the intersection of technology trends and business strategy. His "AutoBeta" piece generalized autoresearch to general-purpose AI work.
Less technical but high reach with media-class executives. Useful for tracking what the enterprise press says about AI capability.
Missing someone?
This is a curated list, not a closed one. New voices land in the practitioner pool every month; the page gets refreshed as the field moves. If you know someone worth following who is not here, open an issue on the companion repo or send a note.