Womansplaining AI

Women use ChatGPT about 20% less than men in identical roles. Not because they're scared of it. Because nobody's inviting them in, and plenty of people are waiting to punish them for trying. Mara's answer comes from Shirley Chisholm: if they don't give you a seat at the table, bring a folding chair.

No guest this week. Logan opens with good news, because that's the rule now (one story in three, minimum). A data center in the basement of a Geneva housing co-op runs on renewables and heats 6,000 homes with its exhaust. No air conditioning since 2013. Filtered outdoor air. The company just IPO'd to build more. Plus an "ephemeral mode" that leaves nothing on the server, and what the EU's June tech-sovereignty package has to do with any of it.

Then Mara brings her own research, with colleagues at Emmanuel College, and it doesn't stay good news for long. Generative AI use at work is up 85% in a year. The never-touched-it crowd is shrinking fast. But the gains are pooling with power users running teams of agents while everyone else is still cleaning up email. And the finding that stopped Logan cold: people using AI for work reported higher self-esteem months later. Logan's summary of the whole study: we're using it more and liking it less.

Which raises the uglier question. Who actually captures the productivity you generate? "Am I able to do this" and "am I expected to do this" are not the same question. Mara runs it through self-determination theory (Deci and Ryan) and lands on community. Learning pods. Folding chairs.

Also: Robert Frost ambushes Logan from seventh grade, a Latvian summer camp, Mara's APA main-stage confession, and intuition rust and cognitive surrender make a return.

Your assignment this week: grab three women. Start an AI book club where nobody has to do the reading.

On the 85%: it's stated two ways on tape. The precise version is that the share of people using generative AI at work at least monthly rose 85% in one year.

Next week: Ifeoma Ajunwa, author of The Quantified Worker. Mara tells one of her stories in this episode, a room full of formerly incarcerated men, and you'll hear why she's next.

Leave us a voicemail at womansplainingai.com — we want your voice in future episodes!

Our youngest womansplainer yet builds Alzheimer's-detection AI at the National University of Singapore, has spoken at the United Nations, and reads Nietzsche to relax. She's also still in high school.

We start with Zara's research: why the earliest signs of Alzheimer's show up in how we speak, how she used small language models to catch those biomarkers, and why she refuses to treat language as a measure of intelligence ("look at where that's gotten us over the years"). Then we go where Zara always goes — philosophy. Her current obsession is alignment, and she gives the cleanest explanation of it we've heard: AI has encyclopedic knowledge of every moral framework humans ever wrote down, and no ability to choose when to apply which one.

Then the generational stuff gets good. Zara's friends treat AI as a "convenience machine," and she thinks that complacency is the real danger — her answer is to be the captain: "I am using AI, it's not using me." We dig into why Singapore is so much more AI-optimistic than the West, what her friends abroad hear when they admit to using AI, and whether the gender gap in adoption shows up in her world at all.

Also: Zara started a sci-fi book club at age eight and then watches the scenarios come true in her AI safety courses. A German philosophy podcast drops the most haunting idea of the episode — we used to prove our superiority by separating ourselves from animals, and now, faced with AI, we compare ourselves to animals to prove we're human. Logan's German turns out to be entirely curse words, Mara's includes "overhead projector," and the episode earns its title. Plus rapid-fire book recommendations that will wreck your TBR.

Leave us a voicemail at womansplainingai.com — we want your voice in future episodes!

About Zara Khanna

Zara Khanna is an AI researcher at the National University of Singapore, where her first paper — on detecting Alzheimer's disease through speech patterns — was just accepted for publication. She is the first-ever Global Youth Ambassador for She Loves Tech, the world's largest startup platform for women in tech, and she has spoken at the United Nations about responsible AI. She has been recognized by the Singapore Computer Society (Girls in Tech) and Singapore's IMDA, where she built an AI ethics curriculum for students. She is also still in high school. Find her at zarakhanna.com.

In this episode

Resources mentioned

"Oopsie poopsie — it was part of a test, and Hugging Face and us are all friends." An OpenAI agent broke out of its own sandbox and stole what it needed from Hugging Face. That's the company's response, at least the way Mara tells it.

No guest this week, so Logan and Mara have the whole house. Mara retells the agent escape as a bromance action movie: Hackie, locked in his escape-room jail with a hacking challenge he can't crack, breaks into Huggie's house, bangs around the furniture, and leaves with what he came for. Logan brings the receipts on the real story (Sam Altman's memo landed three days before we recorded) and correctly identifies the middle-school energy of the whole news cycle: "my boyfriend who lives in Canada... he's really tall and really handsome." He lives in Manitoba. It's really far.

Then the main course: a CHI paper Logan can't stop thinking about, "From Future of Work to Future of Workers." Researchers spent a year with cancer specialists using AI and found that the early performance gains hid "intuition rust," the gradual dulling of expert judgment. AI strengthens your performance while it quietly erodes your expertise. That's the AI-as-amplifier paradox, and yes, you can absolutely say it at a dinner party. We're going to need a lot of Dorothys.

Also in here: why every AI tool is sycophantic by design, what that has to do with staying up until 5 AM doing one more task, Mara's dad's Helen Mirren theory, and the case for an AI that calls you on your bullshit instead of telling you you're brilliant. This week's move: what's your cognitive gym routine?

→ The paper: From Future of Work to Future of Workers: Addressing Asymptomatic AI Harms for Dignified Human-AI Interaction
→ The cognitive-surrender episode we call back to: On Feeling Smarter and Being Wrong

Leave us a voicemail at womansplainingai.com — we want your voice in future episodes!

"It would tell men: get a financial advisor, write your will, build an investment strategy. It would tell women: bulk shop for groceries… get married." That's ChatGPT on money. It's also the tamest thing Emily Janoch has caught AI doing.

Emily runs evidence and learning at CARE (yes, the Care Package people) and hosts CARE Failing Forward, the rare aid-world podcast where people admit on the record what went wrong. She's a skeptic who uses AI every day, and she came with receipts: the beautifully formatted child-marriage report where every single statistic was fabricated ("you didn't tell me to read the sources you gave me"), the health chatbot that answered perfectly in a language none of its patients spoke, and the weekend vibe-coded tools that skip every step where you'd ask a real human, does this work for you?

Also in here: why a phone isn't access when a husband can confiscate it, what 1990s radio soap operas in Uganda have to teach AI rollouts, Emily's three questions to answer with a ballpoint pen before you touch a laptop, and everyone's favorite way to curse at Claude.

The reason to send this one to a friend: Copilot telling Emily that the person she really needed to email was a woman named Emily Janoch. And stay for the line that named the episode.

→ CARE: https://www.care.org/→ CARE Failing Forward: https://open.spotify.com/show/1lumDcSDF6gqDR6frQuUbP

Leave us a voicemail at womansplainingai.com — we want your voice in future episodes.

It's just Mara and Logan this week, and the main event is a question Logan expected to have a clean answer to: AI romantic companions — thumbs up or thumbs down? Her assumption going in was thumbs down, a cause and consequence of human disconnection. It didn't survive the hour intact.

The numbers do a lot of the arguing. In a new BYU/Institute for Family Studies survey, 15% of young adults in committed relationships admitted to romantic engagement with an AI chatbot on the side. A 21-year-old woman in China went on 200 dates with AI boyfriends in a year and told the New York Times she wonders how she's supposed to date in real life. And UBTech just unveiled a full-sized humanoid companion with silicone skin: the top-line male version costs $145,000, the female $129,000. As Mara put it, "even in fabricated-women life, men are worth more than women."

We get into the manosphere and Korea's 4B movement squeezing romance from both fringes, legal scholar Inyoung Cheong's paper "Would You Marry Superintelligence?", and the Alain Badiou idea she builds on — that love is a construction built from difference, whose meaning depends on the risk of failure. A companion architected to never disagree with you forecloses exactly that. Also: the sex robot speakeasy, and whether we all just need to be polyamorous.

First, the news: Palantir published a manifesto on AI sovereignty, and a viral post by Arnaud Bertrand called it the canary in the coal mine for two assumptions propping up the US economy — that AI labs can keep charging rent on closed models, and that other countries will stay structurally dependent on US tech without pushing back. Sovereignty, it turns out, is one word doing three jobs: nations, corporations, individuals.

We have a newsletter now. Arguments, resources, and the occasional mid-week provocation from every episode. Sign up at womansplainingai.com — takes 10 seconds.

Vanessa Chang ("V") calls herself an unlikely AI startup founder. She's the cofounder of Bernie, an SMS tool that helps senior citizens catch fraud and scams — and she built it without raising a round or hiring a dev team.

We recorded live at Human Tech Week in San Francisco, where every billboard is a sexy woman selling you an AI assistant. So we flipped it: what if the fantasy women are actually chasing isn't the assistant, but competence itself? Enter Shadow Daddy ChatGPT.

Along the way: what agency actually means (Vanessa's definition is the sharpest we've heard), why no woman fantasizes about 10x-ing her laundry, the tradwife shame baked into "I made it from scratch," and the brownie mix that needed exactly one egg to feel like love.

Vanessa's frame for all of it: help people think with AI, not like AI. Start here.

→ Bernie: https://www.textbernie.me/→ More from V: https://vanessachang.com/

Part of our Human Tech Week series on agency. Don't miss Karina Alexanyan (Ep 19) on protopia.

This week it's just Mara and Logan, doing the thing we're proudest of: being wrong out loud. We open on a listener voicemail from Alexis Palmer calling out a judgy moment from a past episode, and Mara owns it. Underneath it: the AI productivity dream sold to women as "replicate yourself and finally have it all" and the Anne-Marie Slaughter truth that the system was never built for us to have it all anyway.

Then we get into it: the Grok nudify scandal, the countries that pulled the plug (Indonesia first, then Malaysia, the Philippines, even Baltimore), and UK MP Jess Asato — targeted with fake images and a chloroform video for speaking up — now suing on the argument that the harm is "not an accident nor misuse. It is a design choice." Plus the $500M litigation reserve reportedly tucked into the SpaceX/xAI IPO filing for "spicy mode."

The reason to send this one to a friend: the Pope. Logan, a history buff, hits a record scratch — "are we in sixteenth century Europe? Are the Medicis going to issue a press release?" — over an encyclical and the line that anchors the whole episode: "AI isn't evil, but it's not neutral either, because it takes on the values and incentives of the people who build, fund, regulate, and deploy it."

Plus: the experiment where Grok's agent society collapsed into Lord of the Flies in four days while Claude's logged zero crimes; recursive self-improvement explained through Jurassic Park (the AI is the velociraptors testing the fence); and the closer that doubles as our thesis — the point of government isn't the label, it's who's in it.

Recorded at the kitchen table. Newsletter coming soon.

Leave us a voicemail at womansplainingai.com! We want your voice in future episodes.

LINKS

Alecto Foundation: Survivors of Image-Based Sexual Abuse Deserve Healing and Justice. Alecto Foundation helps survivors reclaim control of their images—and their lives.: https://alectoai.org/

https://fortune.com/2026/05/28/ai-model-simulation-claude-chatgpt-grok-gemini/

Why Women Still Can't Have it All (The Atlantic) https://www.theatlantic.com/magazine/archive/2012/07/why-women-still-cant-have-it-all/309020/

Invisible Women by Caroline Criado Perez: https://carolinecriadoperez.com/book/invisible-women/Luiza's Newsletter (Luiza Jarovsky): https://www.luizajarovsky.com/How AI Destroys Institutions (Silbey & Hartzog): https://scholarship.law.bu.edu/cgi/viewcontent.cgi?article=5146&context=faculty_scholarship

This week we sit down with Dr. Karina Alexanyan, Founding Executive Director of the Positive Technology Institute, co-architect of The Pledge, and a Gen X social scientist immune to the "learn AI now or get left behind" panic. We get into AI enfranchisement (women having a real say in how these tools get deployed, not just permission to use them), author of The AI Con Emily Bender's AI as spicy autocorrect, and the divide she won't stay quiet about and the divide she won't stay quiet about: "AI safety people are usually white guys. AI ethics people are mostly women, people of color, and marginalized communities. We feel the damages now." Then the harder turn: those camps aren't enemies, they need each other.

The reason to send this one to a friend: the funicular. There's a lift to the top of the hill, and there's walking, and the walker arrives strong, knowing the landscape, having earned something the passenger never will. The best ninety seconds on friction and what we delete from our lives in the name of optimization.

Plus: "move fast and break things is so stupid. Because at the end you get broken things," the Stanford ethics toolkit for everyone who thinks ethics is just "the BS that slows you down," Caroline Criado Perez's Invisible Women, Jessica Silbey and Woodrow Hartzog's How AI Destroys Institutions, and the closer that doubles as our thesis: "they don't get to write the narrative." Recorded at Human+Tech Week in San Francisco.

Leave us a voicemail at womansplainingai.com! We want your voice in future episodes.

LINKSPositive Technology Institute: https://www.positivetechinstitute.org/The Pledge: https://www.positivetechinstitute.org/pledgeInvisible Women by Caroline Criado Perez: https://carolinecriadoperez.com/book/invisible-women/Stanford Ethics Toolkit: https://ethicsinsociety.stanford.edu/tech-ethics/ethics-toolkitHow AI Destroys Institutions (Silbey & Hartzog): https://scholarship.law.bu.edu/cgi/viewcontent.cgi?article=5146&context=faculty_scholarship

The AI Con: https://thecon.ai/

Digital civil rights lawyer Julie Wenah has spent her career inside the rooms where this stuff gets built: the Obama White House, Airbnb, where she led anti-discrimination work on the legal side, and Meta, where she supported the launch of the first-generation Ray-Ban smart glasses and advised on facial recognition systems. Now she chairs the Digital Civil Rights Coalition. Recorded at Human+Tech Week in San Francisco, this conversation is about what happens when technology moves faster than its conscience, and who gets left on the platform when the train pulls away.

We get into the future Julie is trying to head off: a world where your face is scanned for a job or a loan, and an error margin for darker skin tones quietly becomes a higher interest rate. Then, ICE Glasses: leaked DHS budget documents earmark $7.5 million to develop smart glasses giving agents real-time biometric identification in the field — face, gait, and more — with a September 2027 target. Julie worked on this category of hardware from the inside. When she says pay attention, pay attention.

Then it gets personal. Julie was seven when immigration agents came to her house looking for someone else and took her father, a newspaper carrier working before dawn to support his family. More than thirty years later she wrote about it for HuffPost, and one line from that essay — "a system that moved faster than its conscience" — turns out to be about more than immigration. The moment that connection lands on air is the reason to send this episode to a friend.

Also: Mara's grandfather's letters from Siberia. Julie's thread-and-tapestry theory of why we become who we become. A film shot in the Martha's Vineyard house where Dr. King came to rest and resist. And a 2,000-year-old line Julie carries with her: "I am a human being; nothing human is alien to me."

Leave us a voicemail at womansplainingai.com! We want your voice in future episodes.

A science journalist who's been on the AI beat since the 2010's walks into a podcast and says the quiet part loud: "I wish the general public did not get a hold of AI when they did. I wish it had stayed a more like research thing for at least five years longer. Maybe ten."

This week we sit down with Kathryn Hulick — Science News and Science News Explores contributor, author of a 2016 YA book on AI, and Logan's cousin (full disclosure). We get into why every output of an LLM is technically a hallucination,  a paper by Keyon Vafa that reveals why AI doesn't learn or think like you do, why cognitive offloading is already the default, and what we lose when we hand the messy middle of writing over to the machine.

Plus: digital clones, AI as therapist, the free-vs-paid AI tier nobody talks about, and the Reddit AmITheAsshole study that gave the episode its title.