Joanna Stern is PROBABLY Not a Robot
This Waveform (WVFRM) podcast episode is a wide-ranging chat where Joanna Stern explains her new book about letting AI run (almost) her whole life for a year, then zooms out into what AI actually changes in real life (healthcare, self-driving cars, wearables, and robots). They also debate Apple's future leadership and Apple Intelligence, plus the privacy vs convenience trade-off that comes with "always-on" AI devices. Finally, Joanna talks about going independent in media, and they end with a fun A-to-Z typing race.
1. Warm intro and the "New Jersey tech media" joke
The episode opens with the vibe of a real in-person hang—Marques says the show is better face-to-face because it's more personal and more fun. Joanna jokes that they're recording basically in her neighborhood in Kearny, New Jersey.
They immediately start riffing on "AI hype statements" people keep repeating—AI will change healthcare, streets, homes, everything—and Joanna pushes back with the question she keeps returning to: what does that actually look like in day-to-day life?
A running gag appears early: they crown themselves the kings of New Jersey tech media.
"The two most powerful people in New Jersey tech media are in this room."
Joanna also holds up her book to camera as a "promotion strategy," joking that it will annoy the editors.
"This is my new way of promoting the book… just holding it up to your face during the entire podcast."
2. The book: "I Am Not a Robot" and what her AI year really was
Joanna explains the book's full idea: she spent a year trying to use AI in as many parts of life as possible—not only chatbots.
She clarifies a key definition: when she says AI, she doesn't mean only "generative AI" (text/image chatbots). She includes things like self-driving cars, medical imaging AI, and even early humanoid robots—basically the broader world of machine learning that people now bundle into the single word "AI."
"I don't define AI as just generative AI. It's not just chatbots."
She jokes that the title is clever but a nightmare for search engines because if you search "I am not a robot," you mostly get CAPTCHAs.
"If you search 'I'm not a robot,' then you just get CAPTCHAs and it's a disaster."
She also plugs her new media company New Things (plus her YouTube channel), positioning herself as not just an author—but someone rebuilding how she does tech journalism.
3. Apple's "new CEO" talk: why a product leader matters
Marques brings up the (already much-discussed on Waveform) topic: John Ternus potentially becoming Apple's next CEO. The big theme here is that Apple led by a product-first executive could feel different than Apple led by an operations/business-focused leader.
Joanna agrees with the general optimism: Apple under a product-focused CEO could communicate product decisions more clearly and thoughtfully.
"It just means that the most senior leader at the company is going to be able to talk a little bit more thoughtfully about these products."
They contrast CEOs who speak mostly in "big picture" terms with CEOs who can get into the "weeds" of the product. Joanna notes that AI company founders are currently the CEOs most deeply connected to their products—because they often built them.
"The closest CEOs to the products are the AI companies because they're the founders."
4. Why she did the AI experiment: testing the hype against real life
Joanna explains her motivation: around late 2024 into 2025, AI product launches and CEO promises hit a peak, and she wanted to "live inside" those promises and see what holds up.
"AI is going to change our lives… and I was like, why don't we just try to live that and see if that's true?"
Her book is deliberately consumer-oriented: it's meant for people using products, not engineers. It's also structured through the year, starting in winter 2025, to capture what it felt like to live through that specific moment of AI acceleration.
A big "human boundary" shows up: she didn't go absolutely 100% AI-maximalist, because that would destroy her personal life.
"If I went full throttle, I would be divorced and I would have lost everything that was important to me."
She returns to the same grounding question repeatedly—when someone says "AI will change X," what does that mean in practice?
"They say AI is going to change healthcare… change our streets… we're going to get humanoid robots in our homes… What does that really mean?"
And she gives a blunt answer about some of the biggest, flashiest promises:
"It actually means nothing because they're really not ready."
5. The model-comparison trap: why AI rankings don't stay stable
Marques mentions he's wanted to do a "best AI model" comparison (GPT vs Gemini vs Claude vs Perplexity, etc.), but he worries it'll become outdated instantly.
Joanna agrees and says that's why she avoids naming specific models much in the book—because books need a longer shelf life than weekly model leaderboards.
"I'm very clear in this book to say I never actually mention model names… because I knew that… it would be out of date."
She shares a funny example of why AI tools can still be unreliable in basic ways: an image generator repeatedly insists it drew five hamsters when the image clearly contains six or seven—an example of AI "confidently" claiming something false (often called hallucination).
" 'Yeah, no, there's five hamsters in this image.' … 'No, there's six hamsters.' … then it's like, 'Seven hamsters.' "
6. The most surprising discovery: AI is already embedded everywhere
One of Joanna's biggest "surprises" isn't a flashy new product—it's realizing how deeply AI is already inside everyday systems, even for people who say they "hate AI."
"The truth is AI is already in your life. There's just no way you're going to be able to say no to it."
Her concrete example: medical imaging. Many people's X-rays, mammograms, ultrasounds are already being supported by AI tools that highlight suspicious patterns for radiologists.
"Many are going to get their X-rays or mammograms… and they're already being read by AI."
She points out something subtle but important: even if you never ride in a robotaxi, self-driving cars can still affect you because your car shares the road with them.
"That AI is affecting your life because your car is driving next to it."
7. Waymo in New Jersey: self-driving cars meet local chaos 🚗
They swap sightings and concerns about Waymo expanding/testing near the New York / New Jersey area. Joanna says she's nervous about autonomous cars on New Jersey highways because the local driving style is… intense.
"I'm a little scared about Waymos on New Jersey highways."
Marques adds that self-driving cars don't just need "rules," they need to adapt to a city's driving culture—California highways are one thing; NYC/NJ includes aggressive taxis, scooters, bikes, tricky intersections, and classic New Jersey road features like jughandles.
Joanna shares a family trip where they went all-in on Waymo rides in Phoenix—calling it their "Waymo fun vacation"—taking around 40 rides so she could notice the little behavioral patterns.
"We took about 40 Waymos… you can pick up all of the little things the cars do."
They also riff on a future where a Waymo might cross multiple regions (NY → NJ → Philly) and need to behave like a "local driver" everywhere—something they think is still years away, but not impossible.
8. Why a book (not just videos): capturing a "moment in time"
Joanna explains why she chose the book format: she compares the current AI era to the early internet era. In the mid-1990s, many people couldn't imagine shopping, mail, and daily life going through the internet—yet it happened. She sees AI at a similar "maybe everything changes" crossroads, full of speculation.
"It feels like it's good to have this piece that will stand the test of time—to either be completely wrong or completely right."
She frames the core premise like this:
"What if machines are a part of every part of the fabric of our lives? … Just like the internet became."
She also wants the book to reach a different audience than hardcore tech followers: she wants tech-savvy people to recommend it to friends/family who are curious but not deep in the weeds.
"Tell their person in their life… 'you should go read this book because it… gives you a really good understanding of what AI could do for you.' "
9. The big trade-off: convenience vs privacy (and "don't make my life a surveillance state")
They dig into a theme that keeps coming up with wearables and AI assistants: the more helpful the tool is, the more data it tends to need. Marques describes it like a sliding scale: maximum convenience often requires maximum access (microphones, cameras, context).
Joanna agrees there's a line everyone draws differently—and companies have to prove the benefit is worth the privacy cost.
"If we're not providing a great tool and utility, people are not going to deal with the privacy trade-off."
Joanna puts the emotional downside plainly:
"I don't want to record everything. I don't want to live in my own surveillance state."
10. The "B" bracelet: outsourcing memory to an always-listening wearable 🎙️
Joanna shows a wearable (she calls it the B)—a simple-looking band with a microphone that can record conversations. She says she wore it most of the year.
Here's the pipeline she describes, step-by-step:
- The bracelet records audio (green indicator = recording).
- The audio goes to the company's cloud (she says Amazon's cloud, since Amazon acquired it).
- The system transcribes it into text.
- The raw audio is deleted.
- She receives:
- a summary
- a list of to-dos extracted from what was said
"It records everything you do… they transcribe it… get rid of the audio… and then I get a summary… and it also gives me to-dos."
Marques connects this to a common behavior: many of us already "outsource memory" to task apps. The difference is that the bracelet does it passively, without you manually typing reminders.
Joanna calls it "another brain," which captures both the appeal and the unease.
"It's basically another brain… I'm outsourcing my memory."
11. Apple, AI, and Siri: "If Apple doesn't give me an LLM… I don't care" (but Siri is painful)
They talk about Apple's positioning: Apple may not be "winning" the AI race in flashy public demos, but Apple's privacy stance can be a differentiator—even if it sometimes means less convenience.
Joanna's hot take is specific: she doesn't need Apple to ship a super-chatty AI personality—she wants Apple to nail the basics.
"If Apple doesn't provide me a large language model to talk to… I don't care."
But she's ruthless about Siri:
"Siri is so atrocious. It is so terrible."
Her everyday example is painfully relatable: she just wants Siri to play NPR news in the morning while getting ready, but it takes multiple attempts and exact phrasing—like we're back in the era of learning "keyword magic" again.
"Why do I have to ask five different ways… and have to perfectly explain the name… that is insane."
They note that in 2026, people shouldn't have to "learn how to talk" to assistants for basic tasks—natural language should work.
12. The "Verified Human" pin: the best AI wearable of the year 😄
Joanna brings out a gag product: a pin that says "Verified Human. I am not a robot." They treat it like a product review, poking fun at failed "AI pin" gadgets from recent tech history.
Marques plays along, giving it a surprisingly glowing "review," mostly because it makes no claims it can't fulfill.
"Battery life seems infinite."
They joke about the "robot test" being whether a humanoid robot is allowed to wear the pin. Joanna reveals the pin-order process is partially automated with an "AI agent" that routes forms into spreadsheets and emails the publisher.
13. Humanoid robots: exciting, creepy, and maybe the wrong shape 🤖
This is one of the densest, most learnable parts of the conversation.
Joanna says she wants humanoid robots to work because it's a category people have dreamed about forever (movies, cartoons). But she's also clear: today's humanoids mostly don't deliver meaningful household utility.
"We want them to work. But they don't."
Marques's skepticism focuses on form factor: a human-shaped robot is fun and familiar (Jetsons, C-3PO), but can be inefficient for real tasks—and also creeps people out because of the uncanny valley (when something looks almost human, but not quite, and it feels unsettling).
"It's cool, but it's also a little bit creepy when we anthropomorphize things that are clearly not human."
"It has to get through that uncanny valley… and I don't know if I'm willing to deal with that."
Joanna shares a key debate among robotics experts:
- Specialized robots that do one job extremely well (more practical)
- Humanoid robots because "the world is built for humans," so human-shaped robots can use human tools/spaces
"There are people that say we should have custom single-utility robots… and then there's a side [that says] the world is built for humans."
Marques pushes the argument further: humans built the world around human limitations, and robots could do better by redesigning systems around robot strengths. His example is perfect: instead of a humanoid robot driving a normal car (still limited by mirrors, blind spots, steering wheel), a true self-driving car is a "robot vehicle" covered in sensors with direct control.
"We as humans built the world around all of the shortcomings of the human form. And we can do better."
14. Laundry folding robots and Moravec's paradox: why "easy for us" is hard for robots 👕
Joanna goes deep on laundry folding robots—she's tested many and even had one in her house. This turns into a practical explanation of Moravec's paradox: tasks that are effortless for humans (like perception + dexterity) can be brutally hard for robots, while tasks that seem "hard" (like math) are easy for computers.
"Things that are really simple for humans are really hard for robots…"
She describes a home robot with two large arms and claw-like grippers—like an arcade claw machine—struggling because fabric is unpredictable: shirts fall differently each time; identifying sleeves/neckline is difficult; and some robots only handle t-shirts.
"Every time a shirt falls it looks different… it doesn't know where the arms are… where the neck is."
They connect this to factory automation too: sometimes a flexible, dangling hose is easy for a human to connect but hard for robots that need perfect positioning.
The core lesson: forcing robots to copy the human way (hands, arms, humanoid posture) may be the expensive path when a non-humanoid mechanism (conveyors, guides, specialized folders) can do better.
15. Chinese humanoid robots: impressive demos, limited usefulness, and a real security question
Joanna mentions she made a video about viral Chinese humanoid robots (like Unitree G1). She says it did well, but also drew anger because she talked about the "they're coming from China" angle.
Her point: people worry about Chinese EVs and phones, but are oddly casual about importing 80-pound humanoid robots.
"We're so worried about Chinese EVs… but yet we're like, 'Oh, let these giant humanoid robots… come into America.' "
She also describes the reality inside her home: the robot can do choreographed moves (dance, kung fu—her kids love it), but otherwise it mostly sits around.
"Other than that, it just like sat in my house doing nothing."
16. The "humanoid robot bet" for 2026: shipped doesn't mean autonomous
Joanna shares a concrete prediction: in 2026, humanoid robots might ship to select customers, but not as fully autonomous, do-everything home helpers.
"They will ship… autonomous? Zero chance. No way."
They joke that if either of them gets a 1X robot in New Jersey, they have to share it and test it together. Marques jokes about remote-controlling it via VR to do chores—highlighting the difference between "teleoperated" (human-controlled) and "autonomous" (robot decides/acts itself).
They also bring up Amazon Astro as an example of a "robot" that exists but doesn't deliver the dream—more like a rolling Alexa/security camera that gets stuck and can't truly help.
"It was just like an Alexa on wheels."
17. Going independent: why Joanna left big institutions for "New Things"
Marques asks about Joanna's move from places like The Verge and The Wall Street Journal to running her own independent operation.
She says the big surprise is how much non-content work appears when you go independent—she references an "octopus" analogy: you don't just make videos, you also do accounting, email, insurance, admin, operations.
"There's a ton of stuff around the video stuff… the inbox, the accounting, the bookkeeping, the insurance…"
She also says: launching a book while starting a new media company is not ideal.
"I'm launching a book in the middle of all this, which I do not recommend."
But she was inspired by creators like Marques, and wanted:
- more direct connection and control with her YouTube audience
- freedom to do "weirder" experiments
- accessibility beyond a wealthy/older publication audience
"I wanted control over the audience… and… the freedom to do even more stuff and weirder stuff."
Her guiding principle for content stays consistent with her reporting style: test things, don't just repeat company claims.
"We test things. We don't just go and talk to the companies making it."
She also doesn't want her whole identity to become "AI YouTube," especially when people feel oversaturated.
"I don't want to only be AI YouTube. I really don't."
18. Content formats: short, long, and the return of "3–5 minute" videos
They briefly discuss how YouTube changed: traditional mid-length videos (around 3–5 minutes) became less common because platforms reward watch time, pushing creators toward longer videos.
Joanna references Casey Neistat telling her Marques handles the shift well, and she argues there's room again for concise videos that don't need to hit 10 minutes just because.
"It doesn't have to be 10 minutes."
19. Rapid-fire round: Magic Mouse loyalty, browsers, and productivity habits
The "rapid fire" becomes… not rapid, because Joanna reveals she's a Magic Mouse superfan and owns two, so she can swap while one charges (since the charging port placement is famously awkward).
"Because when I'm charging, I like to have them in and out."
Marques (and others) roast the mouse design, especially the charging port placement, and joke that maybe Apple's future product CEO will finally fix it.
"Even if that's the only change they make… just move the port."
They then hit a few quick personal-tech questions:
- Laptop vs desktop: Joanna is laptop always, though she has a Mac mini running some tasks.
- Browsers: She uses several, currently Chrome with vertical tabs, and she's been using Perplexity Comet as an "AI browser."
- She confesses she used Microsoft Edge on a Mac for a long time because of a feature called Collections (basically persistent, saved tab groups / project clusters).
"I used Microsoft Edge for a long time… on a Mac."
She also shares a "trust broke" story: an AI browser tool successfully collected images once, then immediately insisted it couldn't collect images, leading her to delete it—highlighting how inconsistent agent-like tools can feel.
On tasks: she uses both pen-and-paper and apps—paper for urgent, same-day execution, and digital tools (Notion-based systems) for longer-range tracking.
On phones: she's currently on an iPhone 17 Pro. If she had to switch away from iPhone, she'd probably pick a Pixel, though she's experimented with foldables and finds them too big for her typical use (especially commuting/texting).
20. A-to-Z typing race and the "kids can't type anymore" realization ⌨️
They finish with Waveform's tradition: guests type A to Z as fast as possible, with three attempts. Joanna improves by the third try and lands at 9.3 seconds.
She jokes the typing test proves she's human—perfectly matching her book title and the "verified human" pin bit.
"I think we proved that I'm a human—pin on it."
This triggers a broader observation: kids increasingly rely on voice and touch interfaces, so they may not learn classic keyboard typing, file structures, or even "how to Google" the way older generations did.
"My kids don't know how to type… mostly they just do voice."
They compare leaderboard times (some guests are unbelievably fast), and Joanna jokes she'll come back to practice—ideally with a humanoid robot to see how fast it can type.
Conclusion
Joanna's core message is simple: AI claims are everywhere, but the only way to understand them is to ask, "What does that really mean in real life?"—and then actually test the tools. Across healthcare, self-driving cars, wearables, and robots, the episode keeps returning to the same tension: usefulness vs trust, especially when "helpful" AI requires deeper access to your life. And on the creator side, Joanna's move to independence mirrors the AI theme too—big promises, real trade-offs, and a lot of behind-the-scenes work to make the future actually function.
Need a summary like this?
Get instant summaries with Harvest