Deep Questions with Cal Newport
Episode 370: Deep Work in the Age of AI
Date: September 15, 2025
Host: Cal Newport
Episode Overview
In this episode, Cal Newport delves into the real-world effectiveness of AI tools for knowledge workers—specifically experienced software developers—exploring whether AI truly accelerates productivity in cognitively demanding “deep work.” Newport analyzes a surprising new study which shows that, contrary to widespread belief, using advanced AI tools actually slowed developers down. Through careful breakdowns and personal insights, Newport uncovers why AI-assisted “cybernetic collaboration” might be counterproductive for deep work, argues for the enduring primacy of intense focus, and fields listener questions about AI, productivity, and technology’s societal impacts.
Key Discussion Points & Insights
1. Background: AI’s Productivity Paradox (00:02)
- Turmoil vs. Reality: Newport introduces ongoing hype around AI’s impact on work. While superintelligent AI may be overhyped, there remain strong claims that “practical” AI tools will significantly enhance productivity, especially for programmers.
- A Surprising Study: Newport discusses the recent July 2025 METR (“Meter”) study entitled “Measuring the impact of early 2025 AI on experienced open source developer productivity.”
Notable Quote:
“It created a glitch in the matrix...leads to some deeper truths about this technology and its potential role in our work today.”
— Cal Newport [00:02]
2. The METR Study: Design and Findings (02:00–06:30)
-
Methodology:
- 16 open-source developers recruited, each working on real and valuable issues (bug fixes, features, refactoring).
- For each issue, use of AI was randomly allowed or disallowed.
- When permitted, developers used state-of-the-art models (like Cursor Pro with Claude 3.5/3.7).
- Metrics: total implementation time self-reported and screen-recorded per task.
-
Expectations vs. Reality:
- Economic and machine-learning experts, as well as the developers themselves, predicted a 20–40% productivity increase with AI.
- Actual Result: On average, programmers were about 20% slower when using AI than when working without it—opposite of predictions.
Notable Quotes:
“The observed result is, on average, they were about 20% slower than the people not using AI... This was an unexpected result.”
— Cal Newport [05:00]
3. Deep Work and the Nature of Programming (06:30–10:00)
- The Concept of Deep Work:
- Programming tasks require prolonged concentration—“deep work”—which Newport argues is the main source of value in knowledge work.
- Supporting tasks (email, meetings, PowerPoints) are not valuable on their own.
Memorable Moment:
“I sort of wrote the book on it. I coined the term. It’s been 10 years, Jesse. Isn’t that hard to believe?...if this book is still selling after 10 years, face tattoo. Boom.”
— Cal Newport, playfully referring to the longevity of his book “Deep Work” [06:49]
4. Cybernetic Collaboration: How AI Was Used (10:00–14:00)
- What Programmers Actually Did:
- Developers embarked on “cybernetic collaboration”—back-and-forth, interactive loops with AI (prompting, reviewing, waiting for AI output).
- Compared to coding solo, less time was spent actively thinking, more time spent managing AI outputs and waiting.
Notable Study Excerpt:
“Developers spend a smaller proportion of their time actively coding … Instead, they spend time reviewing AI outputs, prompting AI systems and waiting for AI generations. Interestingly, they also spend a somewhat higher proportion of their time idle, where their screen recording doesn’t show any activity.”
— METR study, read by Cal Newport [12:15]
Newport’s Framing:
“Let’s call this cybernetic collaboration because these programmers are collaborating on their deep work with a computer...trying to split the cognitive effort...between them and this digital mind.”
— Cal Newport [13:00]
5. Why Cybernetic Collaboration Falls Short (14:00–20:00)
- Breaks vs. Focus:
- Unlike human collaboration (the “whiteboard effect”), which increases focus, AI-assisted collaboration enables frequent breaks and shallower engagement.
- This feels more pleasant but leads to reduced intensity and duration of focus—ultimately reducing productivity and quality.
Notable Quotes:
“When you downshift your mind...let me downshift my focus intensity...it just doesn’t work as well. It might feel nice, but deep work doesn’t really have a lot to do with nice.”
— Cal Newport [18:35]
“Cybernetic collaboration means much less intensity of focus, much less duration of focus. It takes less energy, it feels nicer, but that’s why they’re slower—because intensity of focus is what tells you how fast you’re going to go.”
— Cal Newport [17:50]
6. Takeaways: Deep Work Still Reigns Supreme (20:00–22:00)
- Key Conclusions:
- Anything that reduces intensity of focus is likely to make skilled knowledge workers less productive.
- AI’s current best use: automate shallow tasks to free up more time for true deep work.
- For now, “cybernetic collaboration”—using AI as an interactive assistant for deep tasks—may be pleasant but is unproductive.
Cal’s Summary:
“Deep work rewards intensity of focus. And if you add anything into your workflow that’s going to reduce this intensity, you’ll probably get less productive. This seems to be the trap that a lot of knowledge workers experimenting with AI right now are falling into.”
— Cal Newport [21:15]
7. Listener Q&A Highlights (26:16–31:42)
AI as Project Diaries / Notebooks (26:16)
- Listener “Sam” shares using ChatGPT chats as project journals and asks if it’s an effective practice.
- Newport warns that “cybernetic collaboration” with AI—even for note-keeping—can make work more pleasant but may reduce focus and work quality.
“If it’s reducing your peak intensity of focus or the sustained duration of your focus and you’re doing anything that’s non-trivial, it also could be slowing you down and hurting the quality of your results.”
— Cal Newport [27:03]
Active Recall & Long-term Memory (29:34)
- Newport advises that material not used is forgotten quickly (within ~2 weeks if not actively recalled or applied).
“If you let more than two weeks go without using information at all, it’s going to start to lose its location...The best way to cement something in your memory is to use it.”
— Cal Newport [30:05]
AI and Environmental Impact (31:16)
- Listener “Beth” asks about AI’s environmental footprint relative to everyday tech like Google search.
- Newport notes that ChatGPT-style models are far less energy efficient than traditional search. He suspects current environmental critiques of AI are partly “a way for people who are critical of technology or big tech to get in on…a territory that they're much more comfortable with” [32:48], and that the economics will favor smaller, more efficient AI systems in the long run.
Career Change and Academic Life (38:40)
- Newport discusses his writing, podcasting, academic career, and why he wouldn’t become restless leaving academia.
8. Case Study: Lifestyle Centric Planning (43:29–53:00)
- Newport shares the story of listener “Sven,” who left a tech career to pursue nature-guiding work, only to discover that a move aimed at improving “one aspect” of life inadvertently hurt others (long commute, less family time).
- Sven used Newport’s “lifestyle centric planning” approach to holistically re-calibrate his work and life—returning to tech but deliberately carving out time for nature and family.
Takeaway:
“Your daily subjective mood is not the result of a single decision or change, but on all of the relevant aspects of your life...construct your ideal vision for an entire lifestyle.”
— Cal Newport [44:00]
9. Deep-Dive Mini-Investigation: WiFi Ban and Student Outcomes (53:17–59:15)
- Newport examines claims from a Washington Post op-ed that a West Virginia school (Green Bank) lacking WiFi lags academically because it can’t use modern ed tech.
- His analysis of county- and state-level data finds no evidence that the absence of WiFi caused a unique performance decline compared to similar counties. The root causes of academic underperformance appear more complex than simple access to technology.
Memorable Observation:
“There’s a nice subtle leap from ‘the schools without Wi Fi are worse’ to ‘the schools without Wi Fi are worse because they don’t have Wi Fi’...the picture gets much more murky once you pull even a little bit on the data story.”
— Cal Newport [58:36]
Memorable Quotes & Moments
- “Deep work rewards intensity of focus. And if you add anything into your workflow that’s going to reduce this intensity, you’ll probably get less productive.” [21:15]
- On AI collaboration: “It’s more pleasant, but deep work doesn’t really have a lot to do with nice.” [18:35]
- On focus and productivity: “The brain focusing hard is an incredibly powerful tool. Be wary of things that gets in the way of that...it might be some fool’s gold.” [22:30]
Timestamps by Segment
| Time | Topic/Segment | |----------|-------------------------------------------------------------| | 00:02 | Introduction: AI hype vs. practical impacts | | 02:00 | Overview of the METR programming productivity study | | 05:00 | Presentation of study results – AI makes programmers slower | | 06:30 | The definition and importance of deep work | | 10:00 | Cybernetic collaboration: how developers use AI | | 14:00 | Human vs. AI collaboration: Focus intensity | | 18:35 | Dangers of “pleasant but unproductive” AI collaboration | | 20:00 | Takeaways: Deep work, focus, and the future of AI at work | | 26:16 | Listener Q&A: AI notebooks and project diaries | | 29:34 | Listener Q&A: Active recall and forgetting | | 31:16 | Listener Q&A: AI’s environmental impact | | 38:40 | Listener Q&A: Academic career vs. full-time podcasting | | 43:29 | Case Study: Lifestyle centric planning | | 53:17 | Investigation: Does lack of WiFi hurt student performance? | | 58:36 | Caution against simple cause-and-effect assumptions |
Episode Takeaways
- AI and Deep Work: Using AI collaboratively on deep tasks often reduces productivity and work quality because it encourages less intense, fragmented focus.
- Collaboration vs. Offloading: Human-human collaboration increases focus (the “whiteboard effect”), but AI collaboration so far mostly provides mental breaks, undermining momentum.
- Good AI Use Cases: AI is better positioned to automate “shallow work” and routine tasks, freeing more time for real deep work.
- Productivity Advice: Seek conditions and tools that maximize sustained, intense focus—don’t prioritize pleasantness or ease over cognitive engagement if high-value work is your goal.
- Skepticism on Tech Hype: Be wary of simple cause-and-effect claims about technology, whether in productivity or education—data usually tell a more complex story.
For Listeners Who Didn’t Tune In:
This episode offers a nuanced breakdown of why “cybernetic collaboration” with AI may not be the productivity boon for deep work we’ve been promised. Newport uses both data and relatable analogies to show that, for now, the path to valuable results in knowledge work still flows through sustained, intense human focus—not through comfort or convenience. The lessons here apply to anyone considering how best to use (or not use) AI in their own cognitively demanding pursuits.
