The AI reading list for women who want to understand what’s happening
The most important thing you can do about AI right now is not read more newsletters about it. It's using it every single day. For the work you already do.
The reading list below can help you get up to speed, but reading about AI while not using it risks repeating the mistake many women were pushed into during the early internet era: watching the infrastructure of a new economy get built from the outside. One cost of that mistake was being shut out of a huge share of tech wealth. The cost of repeating it with AI will be higher, and the timeline is that it’s already happening.
If you are serious about being a meaningful part of the next few years, whether as an operator in your field, a founder, an investor, or anything that depends on being current with how the world works, the practice is non-negotiable. Underestimating this technology, or calling it overhyped from the sidelines, is one of the most expensive mistakes you can make.
The reading list is here to help, but it's just the beginning. Whose worth the time?
Simon Willison for capability, Stephanie Palazzolo’s AI Agenda for business, AI Now, Heidy Khlaaf, and CSET for policy and safety, Brian Merchant’s Blood in the Machine for AI and labor, and Mystery AI Hype Theater 3000 for the critical lens.
Twenty minutes a day, alongside the daily practice. That is what it takes to be genuinely informed about AI in 2026.
The longer version is below. Every recommendation needs context, because the goal is not to become a passive consumer of AI commentary. The goal is to build enough judgment to use the technology well, spot what matters early, and understand what it is doing to the world around you.
Why using AI yourself matters more than reading about it
There is a category of professional women right now who are reading three AI newsletters a week, listening to one AI podcast, and have not opened a Frontier AI tool in a month. She is, on paper, informed. She can name the major labs, the latest model releases, and the policy debates.
In practice, she is falling behind every week, because the only way to understand what AI can actually do is to use it on work you care about, fail at it, adjust, and try again.
The reason practice matters more than reading is that AI is not a topic you can fully learn from the outside. It is a tool whose capability ceiling moves constantly, whose useful applications are discovered by users as much as labs, and whose effects on your specific work are invisible until you try to apply it to that work.
Reading about a model’s coding ability tells you something. Spending an afternoon using that model to rewrite your team’s onboarding doc tells you a different thing.
Daily practice does not need to mean building software (though for most people it can). For the reader who has no interest in being an engineer, it can mean drafting emails and documents you would otherwise have written from scratch, then editing the AI version instead of starting from scratch. It can mean using it as a research partner before a meeting where you need to get smart on something outside your expertise. It can mean summarizing long documents you would otherwise skim, working through decisions out loud with it, or replacing one specific task in your week, like a recurring report, meeting prep, or competitive analysis, with an AI workflow and seeing what breaks.
Pick one task this week. Run it through Claude, Gemini, ChatGPT, or another model five times. Notice what works. Notice what does not. That is how the muscle builds.
The reading list below is what helps you make sense of what you are seeing in your own use. It helps you understand the industry forces shaping the tools, the business incentives underneath them, the policy fights around them, and the human consequences of deployment. It’s a support layer for practice.
What’s wrong with the standard AI reading lists
Most AI reading lists solve the wrong problem. They assume the goal is to follow AI. That’s not enough.
The goal is to understand AI well enough to act and take on tasks that previously required a specialized expert. As an operator. As a founder. As an investor. As someone whose field is already being reshaped, whether or not the people around her are ready to admit it.
Millions of people read the same handful of AI newsletters every morning, and most days, many of them cover the same stories. That is not a knock on the newsletters. It is a description of the category. When TLDR AI, The Rundown, Superhuman, The Batch, and The Neuron all summarize the same model released the same day, the marginal value of subscribing to a sixth one is close to zero.
The deeper problem is who gets recommended. Almost every standard “follow AI” list pulls from the same dozen mostly male voices, framing the AI conversation through the founder, investor, and operator lens. That lens is real, and a few of those voices are excellent, but it’s incomplete.
The women doing some of the most rigorous public work on AI capability, safety, policy, labor, and culture are routinely underweighted on those lists.
This is not a list of “women in AI.” It is a list for women who are already using AI, or need to start immediately, and want to understand the field around it.
The best source for understanding what AI can actually do: Simon Willison
Simon Willison’s blog is the single best capability source in 2026, and the most useful supplement to your own daily practice. He co-created Django. He coined the term prompt injection. He builds with new models as they release and writes up what he learns quickly, which is the entire reason the signal is so high.
The reason to read Simon when you are already using AI yourself is simple: he will tell you what the model you used yesterday can do that you probably have not figured out yet. Recent posts have covered agentic engineering patterns, Claude-related security tooling, model behavior, and the Axios npm supply-chain attack. The blog is the primary source. He also publishes a newsletter that functions as an email digest of the blog. Follow whichever format suits you.
A second source worth knowing in this lane is AI as Normal Technology by Sayash Kapoor and Arvind Narayanan, the Princeton computer scientists behind the 2024 book AI Snake Oil. Their thesis is exactly what the name implies: AI is real technology, with real capabilities and real limitations, and should be understood like any other transformative general-purpose technology, not as a singular existential event.
That framing matters because it avoids both mistakes: dismissing AI as hype and treating it as magic. Kapoor and Narayanan are useful because they take AI seriously without accepting the industry’s preferred mythology.
For a research signal without the hype tax, The Batch by Andrew Ng is the efficient option. Light on takes, useful for tracking which research papers and technical shifts actually matter.
The best AI business reporter to follow in 2026: Stephanie Palazzolo
Stephanie Palazzolo writes AI Agenda at The Information, one of the strongest regular sources on AI company internals in 2026. The Information is subscription-only and expensive.
The signal reports on what AI startups are doing: customer counts, revenue figures, hiring patterns, board dynamics, and product realities. This is the information you cannot get by using the tools yourself, but need in order to understand the industry around them.
For startup and VC coverage specifically, Kate Clark at The Wall Street Journal is great. She covers startups, venture capital, and AI. She joined WSJ from Bloomberg in late 2025, after previously helping lead venture and startup coverage at The Information.
Stratechery by Ben Thompson is a source of frameworks. Paid, worth it for the one or two pieces per quarter that change how you think about a question. His aggregation theory has become one of the dominant lenses for understanding how internet companies monetize, and his AI writing is useful when something big happens. You do not need to read every issue. Read him when the market structure changes.
Platformer by Casey Newton is worth knowing about, with one caveat. Newton has been shifting Platformer away from regular newsletter aggregation and toward less-scheduled original reporting and scoops. The publication is still high-signal, but no longer functions as a predictable weekly AI read. Bloomberg’s AI coverage can handle deal flow and earnings.
Skip any AI business newsletter where the author is also an angel investor in the companies being covered. The conflict of interest is endemic in this category.
The best sources on AI policy and military use: AI Now, Heidy Khlaaf, and CSET
This is the lane the standard reading lists handle worst. The framing is usually existential risk: will AI become superintelligent and end humanity? That is a real conversation. It is not the same conversation as what is happening right now, in 2026, to jobs, military operations, immigration enforcement, the legal system, and children.
The smartest voices on the actual effects are often not the loudest voices in the standard AI discourse.
Heidy Khlaaf is chief AI scientist at the AI Now Institute. She is one of the clearest safety-critical voices in AI, with experience evaluating high-risk systems including nuclear power plants, autonomous vehicles, and UAVs, long before “AI safety” became a branding exercise. She also worked at OpenAI on safety methodology for Codex, OpenAI’s code-generation model.
She does not run a personal newsletter. The right way to follow her is through AI Now’s publications, her public commentary, and major interviews when AI hits the news. In her March 2026 PBS interview on military AI use, she warned about the reliability of AI systems used in targeting and pointed to reported Project Maven accuracy rates as low as 30 percent in some situations and average accuracy around 50 percent, “really not far from flipping a coin.”
It is one of the most important public warnings I have seen from an AI researcher this year. The standard “AI to follow” lists do not include her. They should.
For the broader policy lane, follow CSET, Georgetown’s Center for Security and Emerging Technology. CSET is one of the most useful places to track frontier AI policy, U.S.-China competition, national security, export controls, and AI governance.
Helen Toner is worth knowing, but she is not the main feed. She is CSET’s Interim Executive Director and served on OpenAI’s board during the 2023 Sam Altman firing. She testified before the Senate Judiciary Committee on April 22, 2026, on U.S.-China AI competition, frontier AI transparency, IP protection, and distillation concerns.
She publishes a Substack called Rising Tide, but the cadence is intentionally intermittent. Treat it as a high-signal occasional read, not a weekly subscription. Her CSET page collects her think-tank work, and her outside writing appears in publications such as Foreign Affairs. Treat Toner as a person to know. The feed is CSET.
Skip EA-coded existential risk content unless you specifically want to understand that conversation. It is a real conversation, but it's also different from what is happening to jobs and lives right now.
The best newsletter on AI and labor: Brian Merchant’s Blood in the Machine
The labor lens matters because the same AI tools you are learning to use yourself are reshaping industries around you in real time. The clearest source on what that reshaping looks like is Brian Merchant’s Blood in the Machine.
Merchant is a technology journalist and the author of the 2023 book Blood in the Machine, about the original Luddite rebellion. The newsletter takes that historical frame and applies it to the current AI cycle: which CEOs are using AI to consolidate power, which workers are organizing in response, where automation is displacing labor, and where the displacement story is hype.
The cadence is regular enough to follow, the reporting is original, and the framing is the rarest thing in AI media: a labor-first lens rather than a founder-first one.
If you are using AI in your own work, this is the newsletter that will keep you honest about who else that same technology is hurting.
Anne Helen Petersen’s Culture Study is also worth mentioning here for the broader culture-of-work angle, though it is not primarily an AI newsletter. Petersen covers work, parenting, religion, and celebrity, with AI-and-work showing up occasionally rather than as a beat. Culture Study moved from Substack to Patreon in 2025, so subscribe there.
The best podcasts for hearing what AI builders actually think: Hard Fork and Mystery AI Hype Theater 3000
This is the most epistemically dangerous lane, because the builders have incentives to talk their book. Read it with a permanent grain of salt. You still have to read it, because not hearing from the people building the technology is its own form of blindness.
Hard Fork at The New York Times, hosted by Kevin Roose and Casey Newton, is the closest thing to an honest weekly news synthesis. Both hosts are skeptical enough to push back but plugged in enough to keep the conversations going. Start here.
TBPN is useful because it shows how the builder class talks to itself. But it is no longer outside the media. OpenAI acquired TBPN in April 2026, so treat it as a builder narrative rather than an independent analysis. Knowing how OpenAI’s media operation frames the AI story is valuable. Mistaking it for outside analysis is not.
Lenny’s Podcast handles the product side. Lenny Rachitsky interviews product leaders and growth operators, where the actual decisions about how AI is embedded in consumer products are made.
Mystery AI Hype Theater 3000, hosted by Emily M. Bender and Alex Hanna, is the required counterweight. Bender is a linguist at the University of Washington who co-wrote the original “Stochastic Parrots” paper with Timnit Gebru, Angelina McMillan-Major, and Margaret Mitchell. Hanna is director of research at DAIR, the independent AI research institute Gebru founded after leaving Google in 2020.
Bender’s May 2026 piece, “Stochastic Parrots: Frequently Unasked Questions”, revisits how the phrase has been used and misused five years on. The podcast Bender and Hanna host is consistently one of the smartest critical conversations about what AI companies are actually doing versus what they claim. When a lab makes a big announcement, the Bender-Hanna deconstruction usually arrives close behind.
Skip any podcast where the host has an undisclosed financial interest in the AI company being interviewed.
How to actually use this list
The mistake is subscribing to everything and reading nothing.
The bigger mistake is reading and not practicing.
A reasonable starting stack is Simon Willison for capability, Stephanie Palazzolo’s AI Agenda for business if budget allows, Bloomberg if it does not, AI Now, CSET, and Heidy Khlaaf for policy and safety, Brian Merchant’s Blood in the Machine for the labor angle, and Hard Fork plus Mystery AI Hype Theater 3000 for the builder lane.
Five lanes. Twenty minutes a day.
Alongside that reading, pick one professional task this week and run it through a frontier AI tool five times. Notice what works and what does not. Next week, pick another task.
The reading helps you understand what you are seeing in the practice. The practice is what builds the muscle that matters.
Why these women belong on every “follow AI” list, and why reading them is not enough
Helen Toner served on OpenAI’s board during the company's most consequential governance crisis. Heidy Khlaaf worked on safety methodology for OpenAI’s Codex and now does public technical work on military AI reliability, an area far too few people are working on. Emily Bender co-wrote the paper that named the “stochastic parrots” debate. Timnit Gebru helped force the original public fight over the risks of large language models, then founded DAIR to keep doing that work independently. Stephanie Palazzolo and Kate Clark are two of the sharper business reporters covering how AI companies are actually being built and funded.
These are not minor figures. They are people whose work helps explain how AI gets built, governed, funded, sold, and lived with. The standard “follow AI” lists routinely underweight them in favor of louder voices with less depth.
Once you see who is being left off, you cannot unsee the gap.
But also: reading them while not using AI yourself is not enough. The point is not to become the best-informed person standing outside the room. The point is to get inside the work while the work is changing.
