Tech Brew Ride Home
Episode Summary: "How To AI With WSJ's Chris Mims"
Host: Brian McCullough (Morning Brew)
Guest: Christopher Mims (WSJ columnist, author of How to AI: Cut Through the Hype, Master the Basics, Transform Your Work)
Date: February 14, 2026
Episode Overview
This episode features tech journalist and author Christopher Mims, discussing his new book on AI’s real-world impact, how non-experts can integrate AI into their work, misconceptions about artificial intelligence, and why expertise matters more than ever. The conversation is practical, wide-ranging, and philosophical, diving deep into how AI is reshaping work, the importance of critical skepticism, and the future of productivity and personal well-being in the age of intelligent machines.
Key Discussion Points & Insights
1. The Purpose of Mims' Book and Who It's For
- Main Theme: Helping regular professionals make sense of AI beyond the hype.
- Mims describes his book as a field guide for the “other 90%”: people who aren't immersed in Silicon Valley but want to understand and leverage AI practically.
- Quote [01:02] (Chris): “This is the guide for... the other 90% of us... to step back and explain... from first principles, this is what the fundamental architecture is here... and how does that play out in fields where you wouldn't expect... rapid adoption or disruption.”
2. Rethinking "Artificial Intelligence"—Mims’ Preferred Terminology
- Mims prefers "simulated intelligence" to make clear these systems are useful but not sentient.
- Quote [02:50] (Chris): “I call it simulated intelligence because it puts it at a little bit of a remove, it gives us a little bit of skepticism... there’s tons of simulated things that are enormously useful and transformative.”
3. The "Toddler" Analogy – How to Think About Current AI
- AI can be astonishingly capable, but also unpredictable and needs continual guidance.
- Quote [04:09] (Brian): “AI is like dealing with a... super intelligent toddler... you have to keep guiding, prompting, treating AI like a toddler that’s really, really smart, but still has to be guided.”
- Quote [04:56] (Chris): “…Sometimes that toddler will unthinkingly try to cross the street and... you have to yank it back before it nukes your project.”
4. From AI Skeptic to Convert—Mims' Personal Shift
- Mims discusses how he overcame skepticism and found profound uses for AI in his own journalistic workflow.
- Tools like NotebookLM have become essential for synthesizing, summarizing, and note-taking.
- Over-reliance on AI can degrade quality: “if I do too much cognitive offloading, the quality of my work goes down.” [06:32]
- Breakthrough moment: NotebookLM made complex research easier, akin to how Slack democratized messaging.
- Quote [07:27] (Chris): “...when you are handed a tool that solves problems that are essential to your work... that mind blown moment... this thing has always been the most tedious... and now this thing speeds me up.”
5. "Machine Psychology" vs. Trying To Understand the Math
- Advocates for treating AI as a psychological system—probing, observing, and learning its quirks—rather than trying to understand its algorithms.
- Quote [09:43] (Chris): “...machine psychology is my plea to be like, look, the same way that we are all constantly psychoanalyzing ourselves and other human beings, we can apply that skill to... AI.”
6. Mims’ Laws of AI and Why Experts Benefit the Most
- 1st Law: “AI is an assistant, not a replacement.”
- Counterintuitive finding: AI scales the productivity of domain experts, not amateurs (who can’t fact-check or prompt effectively).
- Quote [12:38] (Chris): “If you're an expert, you can do two things with AI that an amateur cannot do. Number one, you can evaluate its work... Number two, you know what questions to ask.”
- Amateurs can produce more surface-level work with AI, but closing the loop and ensuring correctness always requires expertise.
- Quote [14:46] (Chris): “You can create more work-like products as an amateur. But you cannot close the loop and finish that work unless you have the expertise.”
7. Best Starting Point: Give AI Your Least Favorite Tasks
- Begin integrating AI by offloading repetitive, annoying, or low-value tasks.
- Quote [18:03] (Chris): “…such a basic one, but it really is transformative... perfect AI note-takers... can then summarize a meeting at the end are transformative for everybody. Doctors, educators, lawyers—they're all being freed from drudgework.”
8. The Need to Rewire Our Work Habits
- AI tools excel at things we “forget to try” because of old habits. The need is not just for new tools, but for a new mentality of experimentation and willingness to "see what AI can do."
- Quote [22:05] (Chris): “It really rewards experimentation... it has a capability overhang... we would still spend the next decade figuring out new ways to apply them.”
9. Real-World Adoption Examples
- Legal Field:
- Texas personal injury lawyer using Filevine's deposition copilot to ensure all required answers in depositions are actually obtained.
- **Quote [25:05] (Chris):** “It will not check off that you have sufficiently answered that question until... it has heard you get the answer...”
- Clorox:
- AI in product brainstorming led to surprisingly creative product ideas (e.g., “toilet bomb”) by injecting randomness into corporate innovation sessions.
- **Quote [28:04] (Chris):** “...AI can be good at injecting randomness... It can just help give you random ideas...”
- Supply Chain/Prediction:
- Classic (non-generative) AI in supply planning and insurance (Allstate) remains crucial, shows older AI tech is still transformative.
- **Quote [31:05] (Chris):** “A lot of the way they’re killing it with AI is AI from five years ago.”
10. Robotics and the Myth of the “Robot Revolution”
- Despite the hype, we're far from humanoid robots; biggest impact is in incremental improvements (self-driving, logistics, drones).
- Quote [32:11] (Chris): “This idea that we're going to have humanoid robots which are going to have this takeoff moment... remains pretty silly.”
11. Data as the “New Oil”—Who Holds the Power?
- The real power belongs to companies with deep, domain-specific data archives (e.g. LexisNexis, Goldman Sachs, medical/pharma firms), not just big tech.
- Quote [34:04] (Chris): “...LexisNexis with all of their 30 years of database of case law is an incredible example. Goldman Sachs... JP Morgan... If one of the big frontier model companies has licensed some huge volume of data... then it’s them, but none of us know what that is yet.”
12. AI-Fueled Scientific Breakthroughs
- We're already in an era of AI breakthroughs in science: AlphaFold was crucial to rapid COVID vaccine development.
- Quote [37:11] (Chris): “...without AlphaFold... you would not have the COVID mRNA vaccine. That rollout was uniquely enabled by that.”
13. AI, Productivity, and the Risk of Burnout
- The “AI turbocharge” can make work overwhelming, addictive—similar to social media’s slot-machine effect.
- Quote [42:00] (Chris): “...we just keep inventing new ways to hook us on intermittent reward... What I try to do with AI... is make myself into the 0.5 version of myself... instead of a 10x engineer who’s 10 times more productive, what about a 1x engineer who does half as much work?”
- AI can be used not just to do more, but to free ourselves for deep work, daydreaming, or life.
- Quote [45:50] (Chris): “There is so much to be said for using [AI] to take over certain toil so that we can free up that space... because there’s a huge difference between ‘I’m going to use AI to do this work so much faster’ and ‘I’m going to use AI to free myself to know exactly what I should be doing next.’”
Notable Quotes & Moments (with Timestamps)
- [01:02] (Chris): “This is the guide for... the other 90% of us.”
- [02:50] (Chris): “I call it simulated intelligence... it puts it at a little bit of a remove...”
- [04:09] (Brian): “AI is like dealing with a toddler, like a super intelligent toddler...”
- [07:27] (Chris): “...that mind blown moment where it’s like, this thing has always been the most tedious... and now this thing speeds me up.”
- [09:43] (Chris): “Machine psychology is my plea to treat AI like we treat the human brain—by observing and probing its behavior.”
- [12:38] (Chris): “If you're an expert, you can... evaluate its work... you know what questions to ask.”
- [18:03] (Chris): “...perfect AI note-takers which can then summarize a meeting... transformative for everybody.”
- [22:05] (Chris): “It really rewards experimentation. There is a capability overhang... we’ll spend the next decade figuring out new ways to apply them.”
- [25:05] (Chris): (On legal AI assistant): “...it will not check off that you have sufficiently answered that question until... it has heard you get the answer...”
- [28:04] (Chris): “AI can be good at injecting randomness... Just help give you random ideas.”
- [31:05] (Chris): “A lot of the way they're killing it with AI is AI from five years ago.”
- [32:11] (Chris): “This idea that we're going to have humanoid robots... remains pretty silly.”
- [34:04] (Chris): “LexisNexis... all of their 30 years of database of case law is an incredible example.”
- [37:11] (Chris): “...without AlphaFold... you would not have the COVID mRNA vaccine. That rollout was uniquely enabled by that.”
- [42:00] (Chris): “We just keep inventing new ways to hook us on intermittent reward.”
- [43:25] (Chris): “I try to find ways to use AI to make myself into the 0.5 version of myself... liberated from my desk.”
- [45:50] (Chris): “There is so much to be said for using [AI] to take over certain toil so that we can free up that space...”
Section Timestamps
- Main book thesis, who it's for: [01:02]–[02:24]
- Defining "simulated" (not "artificial") intelligence: [02:24]–[04:09]
- AI-toddler analogy: [04:09]–[05:13]
- Adopting AI in personal workflow: [05:13]–[07:07]
- NotebookLM & breakthrough tools: [07:27]–[09:24]
- Machine psychology: [09:43]–[12:00]
- Why experts benefit most: [12:00]–[15:47]
- AI as an assistant, not a replacement: [15:47]–[17:37]
- Best practices: Start with annoying tasks: [17:37]–[19:16]
- Experimentation & mental overhauls: [20:34]–[24:28]
- Incorporation examples: law, Clorox, classic AI: [24:28]–[31:48]
- Robot skepticism: [31:48]–[33:24]
- Data as new oil: [33:24]–[35:49]
- AI & scientific breakthroughs: [36:49]–[38:46]
- Optimism, concern, and social change: [38:46]–[40:25]
- Burnout, dopamine cycles, the 0.5x engineer: [40:25]–[46:56]
- Closing/Philosophical divide: [46:56]–[47:22]
Takeaways
- AI is most effective as an augmentation, not a full replacement, for skilled professionals.
- Start with simplest, most annoying tasks to build AI adoption and efficacy.
- Machine psychology—understanding AI by its behavior—is more accessible than understanding its math.
- Domain-specific data is the next strategic resource in AI development.
- AI can liberate us from drudgery, but could drive burnout if we constantly chase efficiency.
- Future divides may emerge between those who use AI to do more, and those who use it to do less (but better or more freely).
End of Summary.
For a rich, insightful, and often philosophical conversation about AI’s present and future, this episode delivers both practical advice and thought-provoking perspectives.