The Newsletters I Actually Open When They Land

Most AI newsletters say the same three things in a different font.

I unsubscribed from most of mine this year. Here are the nine I still open the moment they land, and what each one is actually good for.

For measurement: Ray Rike and Peter Buchanan , AI to ROI - By Ray Rike and Peter Buchanan

I build the deployments. They quantify what those deployments actually return. One of their September roundups is where I first saw manufacturing’s paid AI adoption cross 60.4%, before it showed up anywhere else I read. If you need a number defensible enough to put in front of a CFO, this is where it comes from first.

For raw spend data: Ara Kharazian , Ramp Economics Lab

Most “AI adoption” numbers come from surveys people answer generously. Ara’s come from actual corporate card transactions across thousands of companies. That’s how he found 80% of OpenAI and Anthropic’s enterprise revenue sitting inside 1% of customers, a number a survey would never have surfaced.

For the view from inside AI drug discovery: Alex Zhavoronkov, PhD, Forever.AI

I know Alex, and he’s one of the few people who can criticize AI drug discovery from inside it. In “The Naked Emperors of AI Drug Discovery” he argues that most of the field still has no real drugs to show. His test is simple: developmental candidates, human data and audited financials, judged over benchmarks. His newest piece looks at three clinical results that point to lower biological age, and it’s where I go for the longevity side without the hype.

For the gap between demo and value: Stuart Winter-Tear, Unhyped AI

Stuart put into one line what I see in almost every project: latent capability and realised value aren’t the same thing. His piece on AI’s capability-realisation problem is the clearest description I’ve read of why a model can do something and a company still doesn’t. He’s now turning that work into a book on management theory for the age of AI, and sharing the argument publicly as he builds it.

For the skeptic’s case: Ed Zitron, Where’s Your Ed At

Wrong on timing for years, still worth reading every week, for the specific signals he names that would actually confirm a slowdown if one started. I’ve built two notes directly off his checklist.

For agent infrastructure, same day: Nate ’s Substack

When a lab ships something technical, Nate has a clear translation out within hours, not days. His read on OpenAI’s Agents API landed the same day as the announcement, and it’s the piece I’d hand a non-technical exec who needs to understand it in five minutes.

For hands-on builder honesty: Ken Yeung, The AI Economy

Ken went quiet for a summer, then came back having actually built something with Claude Code instead of just writing about people who did. That distinction, between covering a tool and shipping with one, changes what a writer notices. I trust the after more than the before.

For AI in the industries most writers skip: AI Wide Open

Most AI writing lives in software and finance. AI Wide Open covers manufacturing, agriculture and transportation, and their latest piece is on manufacturing, which is where a lot of my own work sits. Their 15-minute AI adoption audit is a useful way to find out whether a team’s AI problem is access, ability, trust, workflow or ownership.

For the analyst’s view from inside a hyperscaler: John Hwang , NextWord

Former AWS generative AI architect, now writing about enterprise adoption friction from the vendor side of the table. Useful specifically because most AI commentary comes from either startups selling or consultants implementing, rarely from someone who used to be the one shipping the platform.

---

Nine is already more than most people should subscribe to. If one of these takes a subscriber slot from a newsletter that’s stopped earning it, that’s a fair trade.

Who’s the one newsletter in this space you’d add to this list that I’m missing? I read every reply.