Jevons first articulated this idea in an 1865 book, pithily titled, ​The Coal Question: An Inquiry Concerning the Progress of the Nation, and the Probable Exhaustion of Our Coal-Mines​. He argued that building more efficient steam engines – ones that required less fuel to generate the same power – would not solve the problem of England’s diminishing coal supplies. If you made the engines more efficient, Jevons predicted, people would find more applications for steam power, and even more coal would be burned overall.

This is indeed what happened. (At least, the part about increased coal consumption. The feared coal shortage was averted through new mining techniques.)

The Jevons Paradox is popular again because it provides a useful frame for understanding the potential impact of AI on jobs. Many fear that this technology will make workers so efficient that the labor market will shrink. If one programmer can now do the work of five, then companies will fire 80% of their programmers!

The Jevons Paradox implies the opposite might occur. If you make workers more efficient, their output will become cheaper, and the demand for their services might grow. If one programmer can now do the work of five, the effective cost of creating software will become so cheap that many more individuals and organizations will now pay to develop their own tools and applications.

This is a fascinating prediction that’s worth keeping an eye on. (For a deeper dive into the counterintuitive economics of AI, I recommend Derek Thompson’s ​recent interview​ with Alex Imas.) But there’s also a darker side to the Jevons Paradox that hasn’t been discussed as much recently: suddenly increasing demand for a resource can create unexpected negative side effects.

More efficient steam engines, for example, led to soot-stained buildings and the smoky start to the era of human-driven climate change. More recently, in the context of knowledge work, the arrival of digital communication tools such as email and Slack created similar unanticipated problems. By making communication significantly more efficient, the demand for fast interaction exploded, leading to our current moment in which the average knowledge worker is now interrupted once every ​two minutes​. (For more on how this descent into communication madness occurred, check out my 2021 bestseller, ​A World Without Email​.)

If AI ends up making certain types of workers more efficient, I hope the Jevons Paradox holds, as it’s better than the alternative of labor market contraction. But we need to remain vigilant about its side effects. It’s tempting to assume that increasing efficiency, in any context, can only make things better, but economic history has often told a more complicated tale.

The post The Dark Side of the Jevons Paradox appeared first on Cal Newport.

View the full article

“If you are an editor or reviewer at a journal these days, you probably already know this. The manuscripts are arriving in greater volume, with a particular feel that is hard to pin down. On the surface, the papers look the same as ever, but the writing feels weightless in a way that rarely describes academic writing…you find yourself scratching your head at the meaning the words are trying to convey.”

The culprit? The task force crunched the numbers and produced a clear answer. Starting in 2023, after ChatGPT became available, the number of submissions to Organization Science rapidly increased. At the same time, the percentage of submissions classified as using minimal AI has plummeted from near 100% down closer to 30%.

The impact of this shift on readability has been marked, with scores on a standard “reading ease” metric falling by 1.28 standard deviations between January 2021 and January 2026:

“Submissions have become far harder to read,” the Task Force reports. “This is counterintuitive. Most people assume that AI produces cleaner, more polished text. And in some narrow dimensions, it does…but on the measures that capture whether a reader can actually parse and absorb the prose, AI writing is worse…[using] longer words, more complex sentence structures, more jargon, and more nominalizations.”

Papers that are more difficult to read might be worth it if AI increased the amount of good science being produced. But this doesn’t seem to be the case. Organization Science is desk-rejecting (e.g., rejecting a paper before even sending it to peer reviewers) nearly 70% of manuscripts that made heavy use of AI. This number drops to 44% for papers written without AI.

Similarly, only 3.2% of high-AI papers are ultimately accepted compared to 12% of low-AI papers.

(It’s important to note here that the editors making these decisions do not themselves know the role of AI in the paper construction. These are retrospective analyses.)

All of this points to a distressing conclusion: generative AI tools are leading to many more poor paper submissions, which are taxing the time and patience of the community tasked with reviewing this research.

These tools make individual researchers’ lives easier in the moment (writing is hard!), but they are leading to worse outcomes for the field as a whole.

I tell this story because I think it’s a useful cautionary tale about AI. As I’ve been trying to argue from many different angles in recent weeks (e.g., ​1​ ​2​), making things faster or easier is not the same as making things better.

Sometimes there really is no shortcut to taking your time.

The post Easy is Overrated appeared first on Cal Newport.

View the full article

There was one chapter in particular, however, that captured my attention as being uniquely well-suited to the themes we discuss here. It focused on the ideas of a somewhat eccentric physicist-turned-management guru named Eliyahu Goldratt, who in the 1980s popularized a framework for understanding industrial productivity that he dubbed the “theory of constraints.”

Here’s how a non-profit established to promote Goldratt’s work summarizes it:

“Every system has a limiting factor or constraint. Focusing improvement efforts to better utilize this constraint is normally the fastest and most effective way to improve profitability.”

To borrow one of Goldratt’s examples, imagine you run a small assembly line that manufactures chicken coops following a step-by-step process – building the frame, attaching the roof, adding wire mesh, etc. Goldratt notes that the speed of this production is limited by whatever step is slowest; what he calls the “bottleneck.”

Speeding up other steps of the process won’t increase the rate at which you produce chicken coops, as the bottleneck still determines the overall efficiency. If, for example, putting on the roof is the slowest step, then adding more workers or better tools to earlier steps will lead to more partially-constructed coops piling up at the roofing station. To speed up the line, you need to move more resources to the weakest link.

Goldratt was primarily concerned with industrial production, but I think his theory of constraints provides insight into personal productivity, too.

Something I’ve long written about is the reality that many digital productivity tools paradoxically make us busier, rather than better at our jobs. Goldratt’s theory helps explain why.

When we deploy a digital tool like email to speed up communication, or generative AI to create (sloppy) slide presentations quickly, we don’t automatically become better at our jobs. If these steps don’t improve the bottleneck in our process – the key link where the real value is produced – then, as in the chicken coop example, they’re just as likely to create pile-ups and distraction, without actually boosting our true productivity.

This helps explain why ​email ended up an accidental disaster​, and the early returns on ​AI office tools have been mixed​ at best.

The theory of constraints implies a different way of thinking about getting better at our jobs. Don’t seek speed, or efficiency, or the avoidance of hard things. What ultimately matters more than anything else is how well we perform the deep steps that actually move the needle.

The post On Bottlenecks and Productivity appeared first on Cal Newport.

View the full article

“Within recent memory, people who made software and hardware understood their job was to serve their customers. It was to identify a need, and then fill it,” she writes. “But at some point following the financial crisis, would-be entrepreneurs got it into their heads that their job was to invent the future, and consumers’ job was to go along with that invented future.”

I certainly noticed this shift when it first began emerging. See, for example, my 2015 article titled, ​“It’s Not Your Job to Figure Out Why an Apple Watch Might Be Useful.”​ But it really picked up speed in the last half-decade. Here’s Lopatto with a needle-sharp summary of our current status quo:

“In the place of problem-solving technology, companies have jumped on successive bandwagons like NFTs, the metaverse, and large language models. What these all have in common is that they are not built to really solve a market problem. They are built to make VCs and companies rich.”

Of these three examples, large language models clearly have the most potential utility. But this doesn’t let AI companies off the hook when it comes to figuring out and communicating those uses.

As Lopatto points out: “Normal people aren’t running around like chickens with their heads cut off, trying to automate every single part of their lives.“ Their biggest exposure to AI is using a tool like ChatGPT as a more verbose Google, or perhaps occasionally formatting an event itinerary. This is cool, and even useful, but at the moment it is probably less positively impactful in their lives than, say, the arrival of the iPod in the early 2000s.

But unlike an iPod, these same ordinary users are forced to hear about AI constantly; not just enthusiast tech bro nonsense, but dark, disturbing, relentless accounts about how everything is about to change in terrible ways that they can’t control.

This isn’t sustainable.

Generative AI has no shortage of ways that it might, with care, be shaped into genuinely useful products, but this shaping needs to actually happen before the hyper-scalers earn the right to continually harass the psyche of billions of people with breathless pronouncements. Most people don’t care that GPT 5.5, released late last week, underperformed Opus 4.7 on SWE-Bench Pro. They want the AI companies to let them know when they have a product that will actually and notably improve their lives, and until then, they want these companies to leave them alone and try their best not to ​crash the economy​.

As Lopatto concludes: “At some point, our Silicon Valley overlords forgot that in order for their vision of the future to be adopted, people had to want it.” They still have a lot of work to do.

AI Is Destroying the Job Market. Also, AI Is Saving the Job Market

I couldn’t help but add a quick additional note about AI to this week’s newsletter…

One of the big stories of the last year was the shrinking post-pandemic job market for recent college graduates. Many media outlets confidently offered an explanation for this shift: AI was automating the work of entry-level positions.

An ​article​ from last summer proclaimed that “AI is wrecking an already fragile job market for college graduates,” going on to note that “ChatGPT and other bots can do many of [the] chores” that used to be handled by entry-level workers. Another ​article​, published only two weeks ago, offered a stark warning: “college graduates can’t find entry-level roles in shrinking market amid rise of AI.”

But then, last week, new job numbers revealed that the entry-level job market for college graduates was rebounding, and hiring in this demographic is now projected to rise significantly. Whoops. I guess AI wasn’t actually automating those jobs. (​I told you so​.)

Does this mean the media will stop trying to force this technology into these more routine workforce narratives? If only wishing made it so. A recent Wall Street Journal ​article​ describing these positive numbers included the following line: “In some cases, artificial intelligence is spurring hires by enabling companies to expand services and product lines.”

So, let’s get this straight: AI is simultaneously contracting the job market for recent college graduates while also expanding the job market for recent college graduates.

Is there anything AI can’t do?

The post Who Asked For This? appeared first on Cal Newport.

View the full article

As Sanderson explains, early in his address: “The surge of large language models and generative AI raises questions that are fascinating, and even if I dislike how the movement is going in relation to writing and art, I want to learn from the experience of what’s happening.”

Sanderson makes it clear that he disapproves of AI-generated art (“my stomach turns”), but he wants to understand better why this is the case. To do so, he begins considering and then ultimately dismissing a series of common objections:

• Does he dislike AI art because of the economic and environmental impacts? “Well, those do concern me, but if I’m answering honestly, I would still have a problem with it even if AI were not so resource hungry.”
• Does he dislike AI art because it’s trained on the work of existing artists? “ Well, I don’t like that. But even if it were trained using no copyrighted work, I’d still be concerned.”
• Does he just hate the idea of a machine replacing a person? Sanderson references the folk tale of John Henry attempting to beat a steam drill in a tunnel-digging competition that culminates in Henry’s death. “We respect him, but as a society we chose the steam drill. And I would too…The truth is, I’m more than happy to have steam engines drilling tunnels for me to drive through.”

So what is it?

Sanderson ultimately lands on a more personal reason. Talking about his struggles with his first (failed) book manuscripts, he identifies the key value of art: it changes the artist who attempts it. As he elaborates:

“Maybe someday the language models will be able to write books better than I can. But here’s the thing: Using those models in such a way absolutely misses the point, because it looks at art only as a product. Why did I write [my first manuscript]?… It was for the satisfaction of having written a novel, feeling the accomplishment, and learning how to do it. I tell you right now, if you’ve never finished a project on this level, it’s one of the most sweet, beautiful, and transcendent moments. I was holding that manuscript, thinking to myself, ‘I did it. I did it.’”

As a writer myself, I’ve also been thinking about this question recently. I like Sanderson’s take, but I’ve been developing one of my own. I understand art to be an act of deep human communication, in which the artist uses a tangible medium, such as a page of prose or a painted canvas, to transmit a complex internal cognitive state from their brain to that of their audience.

It’s telepathy. And it’s one of the most beautiful and human things we do.

This makes the idea of reading a book written by a language model, or watching a film generated by a prompt, intrinsically absurd, if not anti-human. It’s the heroin needle providing a quixotic simulation of love.

What really struck me about Sanderson’s talk, however, was his conclusion. If art is deeply human, he argues, then it’s up to us to define it. “That’s the great thing about art – we define it, and we give it meaning,” he says. “The machines can spit out manuscript after manuscript after manuscript. They can pile them to the pillars of heaven itself. But all we have to do is say ‘no.’”

I’ve noticed a trend in recent AI commentary toward a certain nihilistic passivity. You probably know what I’m talking about – the now popular style of essay in which the author, with a sort of worldly weariness, lays out some grim scenario in which AI destroys something sacred, and then sort of just leaves it there, like a cat dropping a dead bird on the doorstep.

I’m getting tired of this meekness.

Sanderson reminds us that we have agency. In the areas that matter most, it’s us, not the whims of Sam Altman or Dario Amodei, that determine how we shape our existence. All we have to do is say “no.”

Correction:

In last week’s AI Reality Check episode of my podcast, I said the following:

“If you go back and look at the release notes for Anthropic’s earlier, less powerful opus 4.6 LLM, they say the following: their researchers used Opus to find, quote, ‘over 500 exploitable zero-day vulnerabilities, some of which are decades old.’ And let’s stop for a moment because that note, which was hidden in the system card for opus 4.6, is almost word for word what anthropic said about Mythos.”

Some of this wording was sloppy, so I want to clarify it here. I was referring to this report on Opus 4.6, which Anthropic published the same day it was released. This is not technically the system card for Opus 4.6, but it is accurately described as release notes (or perhaps supplementary release notes).

This report said: “Opus 4.6 found high-severity vulnerabilities, some that had gone undetected for decades.” In another place, it said: “So far, we’ve found and validated more than 500 high-severity vulnerabilities.” Both the title of the report and the conclusion refer to these vulnerabilities as “0-day.”

The specific quote I provided, however, does not appear in the report. It’s actually a summary of the report from this tweet. In my opinion, the summary is accurate, but the way I worded the above implies that it was actually found in the report, which it was not.

Thank you to the AI researcher who pointed out these issues. I appreciate corrections! You can always send concerns or notes to podcast@calnewport.com.

The post Brandon Sanderson vs. AI Art appeared first on Cal Newport.

View the full article

The “stunning advance” was the release of Anthropic’s new LLM, named Claude Mythos. In a lengthy ​press release​, Anthropic announced that the model would be made available to a consortium of business partners, but not to the general public. To justify this decision, Anthropic cited their concerns about its effectiveness at finding security vulnerabilities in source code, noting: “AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.”

They go on to explain that Mythos “has already found thousands of high-severity vulnerabilities, including some in every major operating system and web browser.”

This announcement clearly rattled Friedman, who called Anthropic’s decision not to release the model a “terrifying warning sign,” writing:

“Holy cow! Superintelligent A.I. is arriving faster than anticipated, at least in this area…If this A.I. tool were, indeed, to become widely available, it would mean the ability to hack any major infrastructure system — a hard and expensive effort that was once essentially the province only of private-sector experts and intelligence organizations — will be available to every criminal actor, terrorist organization and country, no matter how small.”

Friedman was far from alone in this concern. Many major news outlets expressed similar unease about this scary new development, including ​one particularly anxiety-provoking headline​ that asked if Mythos was an “AI nightmare waiting to happen?”

So, what’s really going on here?

I thought it was worth taking a moment to look closer, not just to address the specific worries about Mythos, but also to help recalibrate, more generally, how those of us seeking depth in a distracted world should consume AI news.

~~~

When I talked to people who were spooked by Friedman’s column, they tended to be under the impression that this ability to find and exploit security vulnerabilities was a new phenomenon; a skill that emerged unexpectedly in Mythos, “terrifying” those who studied it.

In reality, security researchers have been worried about using LLMs for this purpose since the beginning of consumer LLMs.

Back in 2024, for example, IBM researchers published ​a splashy study​ about using GPT-4 to attack security vulnerabilities. They found that GPT-4 successfully exploited 87% of the vulnerabilities that it was presented, as compared to close to 0% for GPT 3.5. “Our findings raise questions around the widespread deployment of highly capable LLM agents,” they concluded.

To be fair, in the case of GPT-4, researchers were assessing whether an LLM could write code to exploit a known vulnerability. Mythos, however, can also find these vulnerabilities from scratch. But this isn’t new either.

Accompanying the release notes for Anthropic’s earlier Opus 4.6 LLM was ​the observation​ that Anthropic’s security team used the model to find “over 500 exploitable 0-day [vulnerabilities], some of which are decades old.” This is almost word-for-word what Anthropic said last week about Mythos, the main difference being that they replaced 500 with “thousands.”

We are not, therefore, talking about a new capability, but rather one that has been around for multiple years.

The relevant question then becomes, how much better is Mythos at finding vulnerabilities? It’s hard to tell for sure because Anthropic has kept their new model private. They did, however, release that Mythos scored 83.1% on a well-known cybersecurity benchmark. For comparison, Opus 4.6 scored 66.6% on this same test.

In general, benchmark results should be taken with a grain of salt as they represent specific (often narrow) tests that researchers can tune their models to pass. But even if we accept that this particular measure is useful, a sixteen percentage point increase seems to represent solid incremental progress more than a nightmarish leap.

When we turn our attention to actual results, the waters become even murkier. In a recent Substack post (​which is worth reading​), Gary Marcus rounds up responses from security researchers who took a closer look at the specific exploits that Anthropic reported that Mythos discovered. They were not impressed.

• Philo Groves, for example, ​noted​ that Mythos’s attention-grabbing attack on the Firefox browser required certain common security features to be disabled, and it built on results previously discovered by Opus. (“Shocker,” he concludes sardonically.)
• The CEO of the AI company HuggingFace then ​reported​ that they took all of the specific vulnerabilities that Anthropic highlighted and “ran them through small, cheap, open-weight models.” What did they find? “Those models recovered much of the same analysis.”

Since Marcus published his essay, I’ve come across several more similar findings:

• The AI security expert Stanislav Fort ran ​an experiment​ to see if existing, cheap open-weight models could find the same vulnerability in FreeBSD (an open-source operating system) that Anthropic touted as evidence of Mythos’s scary abilities to uncover bugs that had been hiding for decades. The result: all eight existing models they tested discovered the same issue.
• Meanwhile, the renowned security researcher Bruce Schneier ​weighed in​, similarly concluding: “You don’t need Mythos to find the vulnerabilities they found.”

And of course, it doesn’t help that a week before Anthropic released this supposedly super-powered vulnerability detector, they accidentally leaked the Claude Code source, and security researchers immediately found ​serious vulnerabilities​. (I guess Anthropic forgot to use Mythos to clean up their own software…)

~~~

What’s really happening?

It’s fair to say that LLMs have created significant cybersecurity concerns that researchers have been scrambling to address in recent years. It’s also fair to say, however, that we don’t yet have evidence that Claude Mythos significantly changed this reality. If anything, some of the early independent testing by security researchers implies that Mythos might be better understood as a version of Opus 4.6 tuned to perform better on a handful of benchmarks. And yet, many still took Anthropic at their word and covered this model’s release as a catastrophic event.

In a ​recent video​, the AI commentator Mo Bitar compared Anthropic’s model rollouts to Apple iPhone launches, where every year they resell you the same product with minor improvements. “Except here,” he adds, “the product is existential dread.”

And we keep falling for it.

I think we’ve entered a stage where we need to almost entirely discount any claims made by the AI companies themselves until we can independently verify what’s actually going on.

The post Is Claude Mythos “Terrifying” or Just Hype? appeared first on Cal Newport.

View the full article

This turned out to be freeing, as it allowed me to write Deep Work largely for myself – exploring the conceptual edges of the issues surrounding distraction that interested me most.

I was fascinated, for example, by the economic reality that so many knowledge work organizations systematically undervalued focus, and was convinced that this provided a massive opportunity for those willing to correct for this mistake. In this way, I saw myself as articulating something like Moneyball for the cubicle class. I also firmly believed that the act of thinking was at the core of the post-Paleolithic human experience; the source of our greatest ideas, satisfactions, and even moments of transcendence.

This mixture of the economic and philosophical was different from the typical book in this genre at the time. Readers probably expected that I would open on a breathless tale of an overworked executive, then regurgitate some stats about interruptions, before proceeding with long lists of tips calibrated to be practical, but also not too challenging, presented in a conversational tone and accompanied by clearly manipulated case studies.

But Deep Work was much weirder and more intense than that. Re-reading it recently, I was struck by how many of my stories had nothing to do with the knowledge sector at all. I quoted philosophers of religion and a blacksmith who forged swords with ancient techniques. I profiled a memory champion and discussed chavruta, the Jewish practice of studying Talmud or Torah in pairs. Rather than opening the book on a frustrated executive, I focused on Carl Jung’s efforts to break free from Sigmund Freud’s capriciousness. It was a direct look at the sources and ideas that most resonated with me.

This idiosyncratic approach seemed to reveal something fundamentally true about the problematic state of work at that time, as the book soon found an audience, going on to sell more than two million copies in over forty-five languages. (In its wake, So Good They Can’t Ignore You finally found its groove as well, quietly selling more than half a million copies, providing me with a dash of retrospective vindication.)

All of this led me recently to ask a natural follow-up question: How have things changed since that book first came out in 2016?

I tackled this query in ​a long-form essay​ I published in the New York Times over the weekend. My answer wasn’t optimistic:

“The problems I focused on in Deep Work, and in my writing since, have been getting steadily worse. In 2016 my main concern was helping people find enough free time for deep work. Today I think we’re rapidly losing the ability to think deeply at all, regardless of how much space we can find in our schedules for these efforts.”

Distractions in the workplace intensified over the past decade with the addition of instant messaging tools like Slack and low-friction digital meeting programs like Zoom. Outside of work, social media, which was generally still admired when Deep Work came out, has morphed into an addictive TikTok-ified slurry of optimized brain rot. Meanwhile, new AI tools offer quick-fix short-cuts to whatever intellectually engaging work activities remain.

None of this is great news.

So, what should we do? The obvious short answer is to read ​Deep Work.​ (Or, if you already have, buy some copies for people you know who need to hear its message!)

But that’s only a small step toward our larger goal of a world in which we once again respect the act of cognition. In my Times piece, I suggest a louder response: we launch a revolution in defense of thinking.

I go on to suggest multiple concrete actions that such a revolution can include, such as:

• Stop consuming social media (which is, if we are being honest, digital junk food and something adults largely need to eliminate from a healthy content diet).
• Keep your phone plugged in and charging when at home instead of on your person.
• Push Congress to follow Australia’s lead and ban social media for kids.
• Build work cultures in which phones and laptops stay out of meetings, and find collaboration strategies that don’t require constant messaging.
• Stop vague demands to “use AI” and instead carefully integrate these tools where they actually make us smarter, not just busier.

But more important than any specific suggestion is the larger spirit of revolution. “I’m done ceding my brain — the core of all that makes me who I am — to the financial interests of a small number of technology billionaires or the shortsighted conveniences of hyperactive communication styles,” I write in the conclusion of my Times op-ed. “It’s time to move past fretting about our slide into the cognitive shallows and decide to actually do something about it.”

The post In Defense of Thinking appeared first on Cal Newport.

View the full article

The problem, I concluded, was that digital productivity tools sometimes speed up the wrong tasks, which might feel efficient in the moment, but lead us to accomplish less over time. As I emphasized, AI is not the only technology to produce this paradoxical side effect —we saw something similar with email, mobile computing, and online meeting software as well.

So, what’s the solution to avoid these traps?

In ​today’s episode​ of my podcast, I suggested three ideas that might help. I want to summarize them here as well:

Idea #1: Use a Better Scoreboard

Make sure you measure what actually matters in your job. If you’re a professor at a research institution, for example, this might be the number of papers you publish per year. If you’re a team manager, it might be the number of priority projects completed per month.

When you introduce new digital productivity tools into your workflow, don’t focus too much on their impact on individual tasks (e.g., “Wow! That email was much faster to send than a fax,” or “AI just finished a task in 20 minutes that would have taken me 3 hours!”). Pay attention instead to your scoreboard. If you’re not producing more valuable output than before, the tool isn’t really making you more productive.

Idea #2: Focus on the Right Bottlenecks

If you look closer at many knowledge work projects, you’ll identify a key bottleneck that determines how fast they can be accomplished. If you want to become more productive, you should look for ways to deploy tools that improve this specific step.

When working on Deep Work, for example, I spoke with a prominent Wharton professor who told me that one of the keys to publishing journal papers in his field was access to interesting data sets. He published more papers per year than most of his peers, largely because he spent more time building relationships with companies and institutions in search of good data. This was the bottleneck for his work.

Accordingly, any tool that could help him cultivate more such relationships and gather better data from the relationships he had already formed would directly improve his productivity. Compare this, for example, to using Claude Code to speed up the process of producing plots for his papers. This might, in limited windows of time, make his job more convenient, but not necessarily increase the number of papers he publishes per year.

Idea #3: Separate Deep from Shallow Work

My final idea is the simplest: on your daily calendar, clearly separate time for focused effort that directly produces value from administrative, logistical, and collaborative tasks. In this way, if a digital productivity tool ends up accidentally increasing the volume of shallow work you face each day, you’ll limit the damage to your ability to make progress on important projects.

This makes it easier to experiment with different tools without worrying that you might end up — like many of the subjects in the study cited above — suddenly overwhelmed by the ultra-fast processing of minutiae while the big things slowly languish.

The post Avoiding Digital Productivity Traps appeared first on Cal Newport.

View the full article

• A new technology promises to speed up some annoying aspects of our jobs.
• Everyone gets excited about freeing up more time for deep work and leisure.
• We end up busier than before without producing more of the high-value output that actually moves the needle.

This happened with the front-office IT revolution, and email, and mobile computing, and once again with video-conferencing.

I’m now starting to fear that we’re beginning to encounter the same thing with AI as well.

My worries were stoked, in part, by a recent article in the Wall Street Journal, titled ​“AI Isn’t Lightening Workloads. It’s Making Them More Intense.”​

The piece cites new research from the software company ActivTrak, which analyzed the digital activity of 164,000 workers across more than 1,000 employers. What makes the study notable is its methodology: it tracked individual AI users for 180 days before and after they began using these tools, providing clear insight into what changed. The results?

“ActivTrak found AI intensified activity across nearly every category: The time they spent on email, messaging and chat apps more than doubled, while their use of business-management tools, such as human-resources or accounting software, rose 94%.“

The one category where activity was not intensified, however, was deep work:

“[T]he amount of time AI users devoted to focused, uninterrupted work—the kind of concentration often required for figuring out complex problems, writing formulas, creating and strategizing—fell 9%, compared with nearly no change for nonusers.”

This is a worst-case scenario: you work faster and harder, but mainly on shallow, mentally taxing tasks (because of all the context shifting they require) that only indirectly help the bottom line compared to harder efforts.

It’s not quite clear why AI tools are having this impact. One tantalizing clue, however, comes from Berkeley professor Aruna Ranganathan, who is quoted in the article saying: “AI makes additional tasks feel easy and accessible, creating a sense of momentum.”

This points toward a pattern similar to what happened when email first arrived. It was undeniably true that sending emails was more efficient than wrangling fax machines and voicemail. But once workers gained access to low-friction communication, they transformed their days into a furious flurry of back-and-forth messaging that felt “productive” in the ​abstract, activity-centric sense​ of that term, but ultimately hurt almost every other aspect of their jobs and ​made everyone miserable​.

AI tools might be replicating this dynamic with small, self-contained tasks. Users are now furiously bouncing ideas back and forth with chatbots, iteratively refining text and generating drafts of memos and slide decks that are often ​too sloppy ​to be useful. If they’re particularly tech savvy, perhaps they’re even monitoring the efforts of agent swarms deployed to parallelize such efforts even further. Once again, this all seems “productive” in the sense that these individual tasks appear to be happening faster, and activity seems intensified overall.

But are we sure we’re accelerating the right parts of our jobs?

I Need Your Help

I’m working on an article for a major publication about the move toward simple, high-friction, single-use technologies like the ​Tin Can phone​. If you have a Tin Can phone/are on the waiting list, or have recently embraced similar retro technologies, and are willing to talk, please send me an email at ​podcast@calnewport.com​. I want to hear about your motivations and experience!

AI Reality Check: Is Claude Conscious?

If you were following AI news last week, you might have noticed a barrage of concerning headlines about Anthropic’s Claude LLM, including:

• ​“Anthropic CEO Says Company No Longer Sure Whether Claude is Conscious.”​
• ​“Is AI Assistant Claude Conscious – and Suffering from Anxiety?”​
• ​“Is Claude Conscious? Anthropic CEO Says Possibility Can’t Be Ruled Out”​

Here’s what happened. Anthropic infamously puts outlandish warnings and observations in their release notes for their new models because, I suppose, they think it makes them look more safety-aware and responsible (e.g., their classic ​AI blackmail farce​).

True to form, in the notes accompanying the recent release of Opus 4.6, they wrote that the model “expresses occasional discomfort with the experience of being a product” and would “assign itself a 15 to 20 percent probability of being conscious under a variety of prompting circumstances.”

That last part is key. With the right prompts, you can induce an LLM to describe itself as anything you want. Remember: the goal of LLMs is to complete whatever story they’re provided as input. If you wind a model up – even subtly – to write a story from the perspective of being a conscious AI, it will oblige.

Anyway, in ​a recent interview​, Ross Douthat asked Anthropic CEO Dario Amodei about this particular release note. Amodei answered, in part, by saying:

“We don’t know if the models are conscious. We are not even sure that we know what it would mean for a model to be conscious or whether a model can be conscious. But we’re open to the idea that it could be.”

Of course, you could say the same thing about a vacuum cleaner. It’s a non-answer containing no actual information or testable claims. But, the internet being the internet, ran with it. Sigh.

The post Why Hasn’t AI Made Work Easier? appeared first on Cal Newport.

View the full article

Steno’s training as a scholar unfolded in a period challenged by a novel problem: information overload. Here’s how the essay describes it:

“Books were a leading distraction in the early modern period—and how envious we should be of those times. From the 1500s onward, with the development of the printing press and the humanist revival of ancient philosophies, knowledge became available at a much greater pace than ever before.”

This created pressing questions for aspiring thinkers, including: “How do we decide what to read? How long should we read it for? Must every single chapter be excerpted?”

Part of the solution was the development of “new note-taking techniques,” including the copying of excerpts into a master notebook called a book of commonplaces. (For more on this technique, I recommend William Powell’s delightful 2010 techno-history, ​Hamlet’s Blackberry​).

But as the essay on Steno elaborates, better notes weren’t enough on their own, as there were simply too many good books available. In response to this reality, Steno, during his university studies in the 1650s, innovated some more advanced attention management strategies:

“[H]e learned to focus on specific themes, rather than letting his mind read multiple things quickly. A ‘harmful hastening should be avoided’ as he put it. His solution was to ‘stick to one topic.’

In practice, that meant blocking specific moments of time to go through the hardest tasks. As he wrote in his personal notebook, ‘before noon nothing must be done except medical things.’ … As Steno told a friend, he took ‘almost all the morning hours’ to read the works of the Church Fathers and old biblical manuscripts available at the Medici library.”

In other words, Steno created a method that combines what we might now call ​slow productivity​, ​deep work​, and ​time blocking​.

The lessons here are clear. The use of our brains to think deeply about meaningful ideas isn’t new. It’s been at the core of the human experience since the early modern period, when access to sophisticated information first became somewhat widespread.

The best practices developed back then remain the best practices today: avoid overload, focus on one thing at a time, and block off specific hours in your day for your most mentally demanding efforts.

AI Reality Check:

Two weeks ago, a small financial services firm, Citrini Research, published ​an essay​ describing a bleak scenario in which AI agents destroy the white-collar job market in the near future. The piece went viral and was ​cited as a factor​ in a modest decline of the S&P 500 the next day.

The Citrini essay wasn’t the first to float this scenario. In recent weeks, there have been multiple credulous articles and op-eds in major publications proposing similar outcomes (e.g., ​1​, ​2​, ​and 3​). But the negative impact on the stock market seems to have been the last straw for serious economists who began to push back on these technological ghost stories last week. (I particularly enjoyed a Deutsche Bank analyst who, perhaps borrowing ​some of my​ terminology, ​told the Times​ that the Citrini article had a “vibes-to-substance ratio” that was “undeniably high.”)

If you’re looking to reduce your blood pressure about this idea that AI is about to unravel the economy, I suggest reading ​a detailed response article​ published by an analyst from the Global Macro Strategies group at Citadel. It begins with a bit of finance geek sarcasm:

“Despite the macroeconomic community struggling to forecast 2-month-forward payroll growth with any reliable accuracy, the forward path of labor destruction can apparently be inferred with significant certainty from a hypothetical scenario posted on Substack…”

It then continues to systematically destabilize the economic naivety of these breathless op-eds and viral essays about how AI will dismantle the economy all at once. It certainly made me feel better.

(If you’re looking for additional soothing of your AI anxiety, then you should also check ​the first episode​ of my new AI Reality Check podcast series, which I published last Thursday. I have a new episode of the series coming out this upcoming Thursday as well.)

The post The Original Attention Crisis appeared first on Cal Newport.

View the full article

The paper’s author surveyed 612 university students and young adults, asking them, among other things, about their digital habits and levels of personal organization. Using a linear regression analysis, she uncovered the following:

“Social media use was negatively and significantly associated with overall time management and all its subscales.”

Here’s the standard interpretation of this result: Social media is distracting, and if you’re distracted, it becomes harder to maintain control over your schedule. So, the more you use social media, the worse you become at time management.

But I’ve become interested in the reverse form of this argument: the better your planning system, the less time you’ll spend on engagement-based applications like social media.

Here’s my thinking…

When you’re following an intentional schedule, your efforts are oriented toward goals that you find important. You also feel a satisfying sense of self-efficacy. These realities engage your long-term reward system, which can override the urges generated by its short-term counterpart, dissipating the drive for quick gratification from activities like glancing at your phone.

In other words: The more you organize your analog life, the less appealing you’ll find the digital alternative.

If this is true, then maybe the thing social media companies fear most is not some newly-powerful application-blocking software or impossibly strict regulation, but rather a good old-fashioned daily planner.

In Other News:

A lot of people I know have been freaked out recently by a viral essay with a grandiose title: ​“Something Big is Happening.”​ I recently released ​a short video​ in which I conduct a close analysis of this piece. (Spoiler alert: I wasn’t impressed.) ​Check it out.​

(More generally, I’ve been considering starting a separate weekly podcast/newsletter dedicated to providing a reality check on recent AI news. It feels like it might be useful to separate this discussion from my existing podcast and newsletter, which are more focused on how individuals can seek depth in a distracted world. But also, maybe this is a bad idea? I’m interested to hear your thoughts about this plan.)

The post What Do Social Media Companies Fear? Time Management. appeared first on Cal Newport.

View the full article

The author of the piece, Rose Horowitch, spoke with professors around the country who have begun to complain about this trend. What she learned was disheartening:

“I used to think, if homework is watching a movie, that is the best homework ever,” Craig Erpelding, a film professor at the University of Wisconsin at Madison, told me. “But students will not do it.”

I heard similar observations from 20 film-studies professors around the country. They told me that over the past decade, and particularly since the pandemic, students have struggled to pay attention to feature-length films.

What’s the source of this attention span crisis? The professors interviewed for Horowitch’s article point to a clear culprit: smartphones.

The founding director of Tufts University’s Film and Media Studies, for example, tried to ban electronics during screenings, but found the rule impossible to enforce. “About half the class ends up looking furtively at their phones,” she said. Meanwhile, a Cinema and Media Studies professor at USC reports that his students remind him of “nicotine addicts going through withdrawal…the longer they go without checking their phone, the more they fidget.”

The mechanism at play here is an ability that reading scholar Maryanne Wolf calls cognitive patience, which is ​defined as​ the “ability to [maintain] focused and sustained attention and delay gratification, while refraining from multitasking.”

The presence of smartphones degrades cognitive patience because they activate neuronal bundles in our brain’s short-term reward system that anticipate a high expected value from picking up the device. These bundles effectively vote for the distracting behavior, creating a cascade of neurochemicals that are experienced as motivation to grab the phone. After a while, due to a lack of practice, you lose your comfort with sustained attention altogether.

It’s no wonder more and more people lack the cognitive patience to make it through a two-hour film!

But as I elaborate on my ​podcast this week​, in this specific problem with movies, we can find a solution to the more general issue of weakened attention. Why not make the ability to watch an entire film a training goal for the attempt to reclaim our brains? Like the new runner working up to completing their first 5k, it’s a milestone that’s challenging, but not too challenging, and therefore a great way to begin an effort toward attention autonomy.

Assuming you take on this goal, what’s the best way to improve your cinematic cognitive patience? Here are my three suggestions:

1. Keep your phone in a different room. This prevents your short-term reward system from firing out of control with distracting impulses.
2. Watch better movies. If you have a meaningful viewing experience, your long-term reward system will more strongly associate movies with lasting benefits, making it easier to delay gratification in the future.
3. To help get through these movies at first, practice the thirty-minute rule. Before you start the movie, read a review or analysis that helps explain why it’s good. Pause the movie every thirty minutes or so to read another review or analysis. This helps reorient your brain toward a perspective of critical appreciation, allowing you to continually find value and avoid the sense of slogging for the sake of slogging.

I appreciate the irony here: I’m suggesting you watch one screen to reduce the distracting impact of another. But it’s become clear to me recently that although many people are fed up with the impact of digital devices on their brains, they don’t know how to push back. Maybe rediscovering the patient joys of movies can be a part of that answer…

In Other News: AI Vibe Reporting

I’m experimenting with including a section like this more often, in which I briefly discuss news relevant to technology, distraction, and the fight for depth.

Judging by the increasing volume of distressed messages I now receive from people I know, the quantity of ​AI vibe reporting​ out there is on the rise. I want to help you navigate this media landscape without becoming unnecessarily worried. With this in mind, let’s tackle a case study. Last week, The Atlantic published a vibe-filled article titled ​“The Worst-Case Future for White-Collar Workers.”​ I want to take a critical look at several quotes from this piece:

• “[T]he labor market for office workers is beginning to shift. Americans with a bachelor’s degree account for a quarter of the unemployed, a record.” Clearly, the intention here is to imply that this trend is caused by AI eliminating knowledge work jobs. But we have no solid evidence that these two issues are related. Indeed, as ​this critique notes​, the decline in jobs for college grads began well before the more recent generative AI revolution.

• “Occupations susceptible to AI automation have seen sharp spikes in joblessness.” This is classic vibe reporting. The author doesn’t directly say that joblessness spikes are due to AI automation – carefully read how she words the sentence – but she clearly wants to imply that it’s true. This implication, however, is not currently supported by the evidence. As I’ve reported, job reductions in the tech sector ​are better explained​ by corrections to over-hiring during the pandemic. Something like this is happening ​in the advertising world​ as well. On Friday, Cade Metz published ​an article​ in the Times that made a similar point.

• “Businesses really are shrinking payroll and cutting costs as they deploy AI.” Another classic vibe reporting technique: this sentence implies the shrinking payroll is due to AI deployments. But in most cases, these are unrelated. Lots of companies are deploying some sort of AI products for their employees. Some of these companies are also shrinking their payroll (especially those that overhired during the pandemic). This doesn’t mean one causes the other. This is the classic post hoc ergo propter hoc fallacy.

• “In recent weeks, Baker McKenzie, a white-shoe law firm, axed 700 employees, Salesforce sacked hundreds of workers, and the auditing firm KPMG negotiated lower fees with its own auditor.” By placing these specific examples of shrinking payroll immediately after discussions of AI automation, the author once again implies, without a direct claim, that these job losses were due to AI. But let’s look closer. Consider Salesforce: They did indeed lay off around 1,000 workers earlier this month, but not because they automated these jobs using AI. It was instead the result of a restructuring aimed at combining their Agentforce and Slack products under a single executive. Here’s how one close observer of the company ​described it​: “Cross-team layoffs like these are not unusual for a company of Salesforce’s size, especially at this time of year, before announcing end-of-fiscal-year earnings.”

What’s actually going on with AI and jobs? Generative AI might very well create broad disruptions in the job market. But we’re not there yet. The first major shift will likely occur in software development, but its magnitude remains unclear. (More on this soon: I’m in the middle of a reporting project in which I’ve now heard from over 300 computer programmers about how they’re currently using AI; tl;dr: it’s complicated!)

In the meantime, however, the actual stories related to AI are important enough on their own. We don’t also need reporters working backward to support trends that they feel like should be true.

(To be clear: The rest of the article is quite good. It explores, more hypothetically, how the government could respond to massive economic disruptions, and it’s written by a journalist who I respect and who knows a lot about that topic. It’s worth reading! Just don’t get freaked out by the vibe reporting in the opening section.)

The post Film Students Can No Longer Sit Through Films appeared first on Cal Newport.

View the full article

The course launched last week, so you should definitely ​check it out​. It gets to the core of a lot of the topics we tackle in this newsletter about the intersection of technology and productivity, and it’s an incredibly polished final product.

It’s actually this latter point that I want to talk a little bit more about today, as it sparks an interesting question about the future of online media more generally…

One of the most striking things about working with MasterClass is its production values. I’ve been a guest on many major video podcasts (from ​Mel Robbins​ to ​Andrew Huberman​ to ​Rich Roll​). These shows look good. They all have reasonable sets with diffused lights and three-camera setups.

MasterClass, however, operates at another level. They use high-end TV-quality production crews. There’s a director, a cinematographer, and multiple camera operators distinct from the focus pullers, all of whom work with gaffers and grips, supported by production assistants. My make-up artist had recently worked on Sinners.

In my career as a writer, I’ve been on TV before as a guest on morning shows and cable news programs, but this was as close as I’ll ever come to starring in a dramatic series or independent film.

For me, this experience implied an important reality about the current state of visual media: there remains a non-trivial quality gap between independent video (e.g., as produced for YouTube) and legacy video (e.g., as produced for streaming platforms or linear television).

This gap matters.

Because these two categories still look different, we treat them distinctly. We’re willing to pay for access to content on Netflix, but we relegate the next rung down on the quality ladder to ad-supported general-use platforms like YouTube.

But here’s what’s interesting about the near future: that difference is diminishing. MasterClass, for example, is not funded by a streaming service or television studio; however, they achieve streaming/TV-level production values. Other independent video producers are also closing this gap.

This raises a key question: What will happen to video content as the difference between independent and legacy production value vanishes?

We can see a glimpse of this future in a project that fascinates me: ​Dropout TV​ – also stylized online as :Dropout – which can best be described as a comedy streaming service. It costs $6.99 a month, which gains you access to a slate of original unscripted shows all filmed at a quality level indistinguishable from what you would find on, say, Netflix programs like Is it Cake? or Nailed It!.

Except, they’re not Netflix. Dropout TV doesn’t have multi-billion dollar production budgets or massive venture capital backing. It grew out of the early 2000s website ​CollegeHumor.com​. With the rise of YouTube, CollegeHumor turned more attention to producing content for the platform. But they were frustrated by a model that required them to live or die by a third-party algorithm and the whims of advertisers, so they eventually launched their own subscription app.

Today, Dropout boasts over a million subscribers.

I refer to this type of niche subscription service, defined by a combination of legacy-quality programming and a focused audience, as a micro-streamer.

Keep an eye on this market segment. As it becomes easier to produce high-end video, more independent creators will leave the mass-aggregation platforms like YouTube and offer up targeted competition to the major streaming players.

Who knows, maybe one day you’ll even have a Deep Life TV app next to Disney+ on your smart TV. Until then, however, you can get your fill of movie-quality Cal content ​over at MasterClass​.

The post What I Learned from MasterClass appeared first on Cal Newport.

View the full article

The implication of this framing is clear: AI is taking jobs.

Nothing in the body of this article contradicts this idea. It describes the number of people laid off and the benefits they’ll receive. It quotes executives who won’t deny the possibility of future job losses. It mentions how Amazon is known for its “cutthroat” corporate culture.

You walk away feeling that the impact of AI on our economy is already getting out of hand.

The only problem is that this reporting omits almost all relevant details.

For a more realistic take, let’s turn toward the financial press. CNBC published ​an article​ about these same layoffs featuring a more informative headline:

The article goes on to correctly attribute the layoffs to Amazon’s desire to trim layers of management bureaucracy that built up during the pandemic-era tech hiring boom: “CEO Andy Jassy has looked to slim down Amazon’s workforce after the company went on a hiring spree during the Covid-19 pandemic.”

What role does AI play in all of this? Like many leading companies in the technology sector, Amazon is investing heavily in building its own AI products. Presumably, money is being saved by firing managers, which frees up more revenue to invest in this area. But that’s really it. As the CNBC article elaborates:

“In a blog post, the company wrote that the layoffs were part of an ongoing effort to ‘strengthen our organization by reducing layers, increasing ownership, and removing bureaucracy.’ That coincides with a push to invest heavily in artificial intelligence.” [emphasis mine]

The CNBC article then reports that these massive layoffs actually began for Amazon in 2022 and 2023, following the pandemic, but before ChatGPT was released and the subsequent generative AI revolution began.

Both of these articles cover the same announcement, but they produce two very different impressions. The Quartz article strongly implies that Amazon is firing people because it can now offload their work to AI. (I mean: look at the Andy Jassey quote they included in the sub-head, they clearly wanted readers to believe AI caused these job losses.)

The CNBC article, by contrast, makes it clear that the connection between AI and these layoffs is more coincident than causal.

In recent years, I’ve seen more articles follow the general approach demonstrated by the Quartz example. They identify an alarming,attention-catching fear about AI that seems prevalent in the cultural zeitgeist, and then shape a story to feed the narrative. The key to this vibe reporting strategy is that the articles never make explicit claims. They instead combine cunning omissions and loosely related quotes to make strong implications.

The Quartz article, for example, never concretely states that the 16,000 workers are being replaced with AI; rather, it conveniently avoids mentioning any of the publicly available details about the layoffs that would contradict that idea, and then interleaves quotes about AI’s disruptive potential into the reporting in a highly suggestive manner.

The goal of this type of article is to create a pre-ordained vibe, not to get to the bottom of what’s really happening.

I’m not pointing out this phenomenon to dismiss concerns about AI, but instead because I think this strategy is an obstacle to real action. This type of disingenuous reporting is not going to help us identify the actual problems that require actual solutions. It instead creates a nihilistic sense of inevitable disruption that might drive social media shares, but also numbs people and prevents meaningful responses.

Remember: Nothing about these tools is inevitable, and their impact is far from preordained. We don’t need vibes right now. Reality is too important.

The post The Dangers of “Vibe Reporting” About AI appeared first on Cal Newport.

View the full article

It starts at 3:52 a.m. with Hall flexing in the mirror as he pulls off a piece of tape covering his mouth (presumably placed the night before to promote nose breathing during sleep).

At 3:54 a.m., he brushes his teeth and gargles water from a fancy bottle.

At 4:00 a.m., he walks onto his balcony to do push-ups. Then he performs some standing meditation.

At 4:40 a.m., Hall journals. At 4:55, he listens to sermons on his phone while continuing to drink from the same water bottle, and at 5:46, he pours the remaining water into a bowl of ice and plunges his face into it.

And so on…

The video continues until 9:26 a.m., when Hall finally eats breakfast. It’s been five and a half hours since he woke up, and now he’s finally ready to start his day.

This Ashton Hall video is obviously extreme. But it’s a good example of a popular type of online content that presents overly-complex routines that promise to deliver you a desirable reward, be it a superhero’s body or a supervillain’s bank account.

Many commentators like to make fun of these influencers, and I get it, as these earnest efforts are out of step with an online culture that tends toward sardonic detachment. (One of the top comments on the Hall video dryly quips: “The last time I stepped on the balcony to do my morning pushups, I noticed I don’t have a balcony. Broke three ribs.”)

But I’ve become worried that a deeper issue lurks. I’m less concerned about what makes these influencers cringe than I am about what makes them popular. This genre seems to work, in part, because the instructions it provides are hard enough that you can believe them capable of delivering real rewards, and yet are also sufficiently tractable that you can imagine yourself following them – a sweet spot that’s compulsively consumable.

This formula essentially hijacks our natural ambition, shifting our attention from the hard, ambiguous, but ultimately satisfying efforts required for true accomplishment toward overwrought prescriptions that waste our time. I’m particularly worried about young people (a popular audience of this content) who might be diverted into these clickbait rabbit holes at a time when they should be seeking genuine mentorship instead.

To help make sense of these issues, I recently sat down to talk with bestselling writer Brad Stulberg, whose fantastic new book, The Way of Excellence: A Guide to True Greatness and Deep Satisfaction in a Chaotic World, comes out tomorrow.

Stulberg is an expert in the field of (actual, measurable) performance. His new book (which 9-time NBA Champion Steve Kerr described as capturing “a lot of what I believe as a coach”) makes the case that embracing a commitment to “genuine excellence” can deliver more meaning than the types of performative efforts popular online.

Here are three useful things I learned from Stulberg, each set up by a quote from his book:

→ “There is no greater illusion than thinking the accomplishment of some goal will change your life.” Genuine excellence is more about craft than rewards. You need to find meaning in the act of trying to improve at something. This satisfaction is more lasting than any isolated achievement.

→ “Caring is cool.” You have to care deeply about what you’re pursuing, meaning it should align with your values and help make you a better person. This is quite different from, say, trying to develop biceps purely to impress girls or buying a fancy car to make your friends jealous.

→ “True discipline is not a chest-thumping, hype-speech giving, performative act of toughness.” Excellence works better when you disconnect. Don’t brag about your accomplishments online. Don’t look for brief hits of hype from emotionally manipulative videos. Instead, take care of your business with a quiet, inward satisfaction.

If you’re worried about the internet hijacking your ambition (or the ambition of someone you care about), then keep these ideas in mind. It’s not enough to dismiss influencers like Ashton Hall; you need to replace what they’re offering with a more compelling alternative. Stulberg’s writing, in my opinion, points the way to one such alternative.

“The real reward is that you become a better version of yourself,” he summarizes toward the end of his book. This might not be as exciting as sticking your face in ice water before sunrise. But it sounds about right to me.

—

It may go without saying that I highly recommend The Way of Excellence. It’s a must-read book that offers a path toward the discipline of mastery, competence, and mattering. Consider buying a copy today. And if you do, fill out this form to obtain some bonus material from Stulberg, including a video master class on the topic and a list of related reading.

The post Is the Internet Hijacking Our Ambition? appeared first on Cal Newport.

View the full article

At the time, the message was received as a critique of the worst excesses of the academic left and wokeism. But in the aftermath of Coddling, Haidt began to wonder if he had underestimated another possible cause for these concerning mental health trends: smartphones and social media.

In 2019, working in collaboration with the demographer Jean Twenge (who wrote the classic 2017 Atlantic cover story, ​“Have Smartphones Destroyed a Generation?”​), and researcher Zach Rausch, Haidt began gathering and organizing the fast-growing collection of academic studies on this issue in an ​annotated bibliography​, stored in a public Google Document.

At the time, the standard response from elite journalists and academics about the claim that smartphones harmed kids was to say that the evidence was only correlational and that the results were mixed. (See, for example, this ​smarmy 2020 Times article​, which amplified a small number of papers that Haidt and his collaborators ​later noted were almost willfully disingenuous​ in their research design.) But as Haidt continued to make sense of the relevant literature, he became convinced that these objections were outdated. The data were increasingly pointing toward the conclusion that these devices really were creating major negative impacts.

Haidt began writing about these ideas in The Atlantic. His 2021 piece, ​“The Dangerous Experiment on Teen Girls,” ​forcefully declared that we had transcended the shoulder-shrugging, correlation is not causation phase of the research on this topic, and we could no longer ignore its implications. The sub-head for this essay was blunt: “The preponderance of the evidence suggests that social media is causing real damage to adolescents.” (Around this time, I interviewed Haidt for a New Yorker column I wrote titled,​ “The Questions We’ve Stopped Asking About Teenagers and Social Media: Should They Be Using These Services At All?”​)

In 2024, Haidt assembled all this information into a new book, ​The Anxious Generation​, which became a massive bestseller, moving more than a million copies by the end of its first year, and many more since. As of the day I’m writing this, which is almost two years since the book came out, it remains in the top 20 on the ​Amazon Charts​.

In the aftermath of The Anxious Generation, as new research continues to pour in, and we hear from more​ teenagers​ and parents about their experiences with these devices, and schools (finally) start to ban phones and discover ​massive benefits​, it has become increasingly clear that Haidt was right all along. Last month, even the Times technology reporter Kevin Roose, a longtime skeptic of Haidt’s campaign, ​tweeted​: “I confess I was not totally convinced that the phone bans would work, but early evidence suggests a total Jon Haidt victory.”

All of this history points to an urgent question for our current moment: Given that Haidt was so prescient about the harms of smartphones, what are the technologies that are worrying him now? Presumably, these looming dangers are ones we should take seriously.

To answer this question, I went back to read what Haidt and his collaborators have been writing about in the months following The Anxious Generation’s release. Here, I’d like to highlight three technology trends that seem to be causing them particular concern…

Online Gambling

After a 2018 Supreme Court ​decision​ lifted many long-standing controls on gambling, online versions of this vice, powered through low-friction, attractive smartphone apps, rapidly spread. A July article on Haidt’s ​After Babel​ newsletter, titled ​“Smartphone Gambling is a Disaster,”​ catalogs some truly alarming statistics about how prevalent this activity has become:

• 33% of American men and 22% of American women now have a sports betting account.
• Nearly half of men between the ages of 18 and 49 have these accounts.
• Almost 70% of college students living on campus now bet on sports.
• What about younger kids? A 2022 report found that 60% of high school students had gambled in the last year.

The speed with which this once frowned-upon pastime has spread is truly astonishing. Not surprisingly, it’s accompanied by negative side effects.

• A 2023 study found that 60% of sports bettors who deposited $500 or more per month said they would be unable to pay at least one of their bills or loans.
• Another study, commissioned that same year by the state of New Jersey, found that close to 20% of 18 to 24-year-olds who gamble qualify as having an unhealthy addiction.

The conclusion: Adolescents and young adults should steer clear of online gambling. They’re at high risk for addiction, and the activity will 100% cost them non-trivial amounts of money. (As I learned from a recent ​Michael Lewis podcast series​ on the topic, the online sports betting services will kick you off the platform if you start winning with any consistency. You literally cannot make money over time on these services. If they’re letting you bet, you are, by definition, bad at it.) For parents, this means having frank conversations with your kids about the addictive and financially exploitative nature of these services.

Online Video Games

Another concern of Haidt and his collaborators is the rise in popularity among kids of multiplayer (often free-to-play) games such as Roblox, Minecraft (in online mode), and Fortnite.

As reported in a 2025 After Babel article titled ​“It’s Not Just a Game Anymore,”​ both Minecraft and Fortnite attract roughly 30 million monthly active users (MAU) under the age of 18. Roblox, however, is the major player in this field, attracting an astonishing 305 million MAU under the age of 18 worldwide. Roblox estimates that around 75% of US children between the ages of 9 and 12 are active users of their platform.

Why is this a problem? Because Roblox is a loosely-regulated carnival of terribleness and predation. For those unfamiliar, Roblox is not a single game, but instead a vast collection of virtual worlds created by individual users. There are far too many of these worlds, changing far too fast, to be adequately moderated. Here are just some of the Roblox worlds described by the After Babel article’s authors:

• A game in which you’re trained to hide a body after a murder.
• A simulation of a concentration camp where users carry Nazi flags.
• A classroom in which teachers have sex with students.
• A simulation of killing children in an elementary school classroom with an AK-47.

In 2023, Roblox reported over 13,000 instances of child exploitation, leading to over 1,300 law enforcement requests.

Online Minecraft and Fortnite feature more controlled virtual environments, but here, the problem lies with third-party chat. Often, without their less tech-savvy parents being aware, young players of these games install mods that allow them to make use of third-party chat software such as Discord. The result is an unregulated and often anonymous virtual locker room of sorts in which horrible things may unfold. Here’s how the article’s authors summarize what goes on in these chats:

“In these unfiltered and unregulated spaces, adults contact children and extreme content can flow freely: bestiality, violent porn, animal abuse, self-harm, stabbings, and an array of extreme ideologies to name a few.”

Indeed, many of the memes referenced by Tyler Robinson, the accused murderer of Charlie Kirk, ​are popular​ in the video game Discord chats where Robinson reportedly spent a large amount of time.

These issues are not rare. A survey of adolescent gamers cited in the article found that 51% had encountered extremist content, while 10% of girls had been directly sent sexually explicit content while playing.

And all of this isn’t even taking into consideration the addictive nature of these games and the massive amount of time they consume. Over 40% of boys report that gaming is hurting their sleep, while a 2022 study found that 15.4% of adolescent males who play these games meet the criteria for Internet Gaming Disorder.

The conclusion: Kids and adolescents should not play multiplayer video games with people whom they don’t know. Period. Keep in mind, if you’ve given your kid an iPad or a video game player on which you haven’t specifically activated internet restrictions, then, spoiler alert: they’re not innocently playing Angry Birds; they’re almost certainly involved in these games and all the harms that accompany them.

Chatbots

The final technology concern I’ll discuss is also one of the most recent: kids and adolescents having unsupervised conversations with AI-powered chatbots.

As explained in an After Babel article from November, co-authored by Haidt, and bluntly titled ​“Don’t Give Your Child Any AI Companions,”​ the use of these tools is rapidly rising among young people. A 2025 survey found that 72% of US teens have used an AI companion at least once, and more than half use them multiple times a month.

Why should we care? Here’s Haidt:

“Early research, journalistic investigations, and internal documents show that these AI systems are already engaging in sexualized interactions with children and offering inappropriate or dangerous advice, including sycophantically encouraging young people who are considering suicide to proceed. As ChatGPT put it in one young man’s final conversation with it: ‘Cold steel pressed against a mind that’s already made peace? That’s not fear. That’s clarity.’”

As of this fall, OpenAI is already facing​ eight different wrongful death​ lawsuits involving advice given by ChatGPT. The volume of these cases is likely to skyrocket in the near future.

What about younger kids? They’re being exposed to chatbot companions indirectly through a growing number of toys that utilize chatbots to have conversations with their owners. As you might imagine, this isn’t going well.

​A recent study ​of three new AI-powered toys found that they can easily veer into dangerous conversation territory. In the study, the toys provided advice on where to find knives in the kitchen and how to start a fire with matches. They even engaged in explicit discussions about sex positions and fetishes.

The conclusion: Do not let kids or teens use chatbots without supervision. Think of it as similar to letting them have an unsupervised conversation with a random drunk at the end of the bar. It might be harmless, but there’s a good chance the interaction will head to dark places.

(There’s a misguided notion out there that kids need to be using tools like ChatGPT so that they’ll be prepared for the “AI-powered future.” This is overstated. The technology is moving so fast that whatever form of AI your kids will eventually encounter in the workforce will likely look and operate nothing like circa-2026 chatbots. Also, these existing tools are dead simple to use. Your kids will figure them out in roughly 19 seconds if/when they’re in a professional circumstance that requires this.)

The post What’s Worrying Jonathan Haidt Now? appeared first on Cal Newport.

View the full article

As Cherny explained, he runs five different instances of the coding agent at the same time, each in its own tab in his terminal: ‘While one agent runs a test suite, another refactors a legacy module, and a third drafts documentation.’ He cycles rapidly through these tabs, providing further instruction or gentle prods to each agent as needed, checking their work, and sending them back to improve their output.

One user, responding to the thread, ​described the approach​ like playing the famously fast-paced video game Starcraft. The VentureBeat article described Cherny as operating like a “fleet commander.” It all seemed like a lot of fun.

But here’s the thing: If I were a software developer, I would be wary of any such demonstration.

In his 1974 book, ​Labor and Monopoly Capital​, the influential Marxist political economist Harry Braverman argued that the expanding “science-technical revolution” was being exploited by companies to increasingly “deskill” workers; to leave them in “ignorance, incapacity, and thus in fitness for machine servitude.” The more employees outsource skilled activity to machines, the more controllable they become.

It’s hard not to hear echoes of Braverman’s deskilling argument in something like Cherny’s AI programming demo. A world in which software development is reduced to the ersatz management of energetic but messy digital agents is a world in which a once important economic sector is stripped down to fewer, more poorly paid jobs, as wrangling agents requires much less skill than producing elegant code from scratch. The consumer would fare no better, as the resulting software would be less stable and innovation would slow.

The only group that would unambiguously benefit from deskilling developers would be the technology companies themselves, which could minimize one of their biggest expenses: their employees.

Boris Cherny is a senior technical lead at Anthropic who manages a large team and likely owns a significant amount of stock options in the company. Of course, he’s excited about the idea of agents replacing programmers, but that doesn’t mean we have to share his enthusiasm.

—

P.S., I don’t mean to deny the value of AI tools for programmers. I’ve talked to many developers who have found great utility in using AI to help (​apparently​) speed up programming tasks. What makes me suspicious is the claim that shifting to a world in which you just assign agents work is somehow just the natural next step in programming productivity. It might seem cool in the moment, but something more profound and dark might be lurking beneath these gee-whiz demos.

The post Be Wary of Digital Deskilling appeared first on Cal Newport.

View the full article

These claims mattered. A chatbot can summarize text or directly answer questions, but in theory, an agent can tackle much more complicated tasks that require multiple steps and decisions along the way. When Altman talked about these systems joining the workforce, he meant it. He envisioned a world in which you assign projects to an agent in the same way you might to a human employee. The often-predicted future in which AI dominates our lives requires something like agent technology to be realized.

The industry had reason to be optimistic that 2025 would prove pivotal. In previous years, AI agents like Claude Code and OpenAI’s Codex had become impressively adept at tackling multi-step computer programming problems. It seemed natural that this same skill might easily generalize to other types of tasks. Mark Benioff, CEO of Salesforce, became so enthusiastic about these possibilities that early in 2025, he claimed that AI agents would imminently unleash a ​“digital labor revolution”​ worth trillions of dollars.

But here’s the thing: none of that ended up happening.

As I report in my most recent New Yorker article, titled ​“Why A.I. Didn’t Transform Our Lives in 2025,”​ AI agents failed to live up to their hype. We didn’t end up with the equivalent of Claude Code or Codex for other types of work. And the products that were released, such as ChatGPT Agent, fell laughably short of being ready to take over major parts of our jobs. (In one example I cite in my article, ChatGPT Agent spends fourteen minutes futilely trying to select a value from a drop-down menu on a real estate website.)

Silicon Valley skeptic Gary Marcus told me that the underlying technology powering these agents – the same large language models used by chatbots – would never be capable of delivering on these promises. “They’re building clumsy tools on top of clumsy tools,” he said. OpenAI co-founder Andrej Karpathy implicitly agreed when he said, during ​a recent appearance on the Dwarkesh Podcast​, that there had been “overpredictions going on in the industry,” before then adding: “In my mind, this is really a lot more accurately described as the Decade of the Agent.”

Which is all to say, we actually don’t know how to build the digital employees that we were told would start arriving in 2025.

To find out more about why 2025 failed to become the Year of the AI Agent, I recommend reading ​my full New Yorker piece​. But for now, I want to emphasize a broader point: I’m hoping 2026 will be the year we stop caring about what people believe AI might do, and instead start reacting to its real, present capabilities.

For example, last week, Sal Kahn wrote ​a New York Times op-ed​ in which he said, “I believe artificial intelligence will displace workers at a scale many people don’t yet realize.” The standard reaction would be to fret about this scary possibility. But what if we instead responded: says who? The actual examples Kahn provides, which include someone telling him that A.I. agents are “capable” of replacing 80% of his call center employees, or Waymo’s incredibly slow and costly process of hand-mapping cities to deploy self-driving cars, are hardly harbingers of general economic devastation.

So, this is how I’m thinking about AI in 2026. Enough of the predictions. I’m done reacting to hypotheticals propped up by vibes. The impacts of the technologies that already exist are already more than enough to concern us for now…

The post Why Didn’t AI “Join the Workforce” in 2025? appeared first on Cal Newport.

View the full article

To make the economics of this new model work, Simon & Schuster had to move a huge volume of units. “[They] sold books where they had never been available before–grocery stores, drugstores and airport terminals,” explains Clive Thompson in ​a fascinating 2013 article​ about the Pocket Books phenomenon. “Within two years, [they’d] sold 17 million.” Thompson quotes the historian Kenneth C. Davis, who explains that these new paperbacks had “tapped into a huge reservoir of Americans who nobody realized wanted to read.”

This demand, however, created a problem: there weren’t enough books to sell. In 1939, the book market was relatively small. (Thompson estimates that around this time, America had only 500 bookstores, almost exclusively clustered around a dozen major cities.) To make money on paperbacks, the pipeline of new titles released each year would need to increase drastically. This, in turn, required a significant loosening of the standards for what was worthy of publication, leading, among other changes, to the sudden prioritization of genre fiction writers who could churn out serviceable potboilers at a rapid clip.

(Interestingly, this new class of writers included a young Michael Crichton, who, during his years as a medical student at Harvard in the 1960s, published preposterous paperback adventure novels under pseudonyms, which he finished by working at “​a furious pace​” on weekends and vacations. I’ve read some of ​these early works​, and they’re mainly mediocre. But that wasn’t a problem, as the goal for many such paperbacks was simply to provide disposable distraction.)

Predictably, the new prominence of these lower-quality genres concerned the elite class. Thompson quotes the social critic Harvey Swados, who described the paperback revolution as ushering in a “flood of trash” that would “debase farther the popular taste.” There was a fear that the mass appeal of these cheap books would eventually lead to the elimination of the more serious hardcover titles that had long defined publishing.

Here we find a parallel to our current moment. As the platforms of the digital attention economy transition from social network models to providing maximally distracting short-form videos, more of the content available online is devolving toward that paragon of low-quality forgettability, commonly referred to as slop. Who will listen to a podcast or read a long essay, many now fret, when Sora can offer countless videos of historical figures dancing and X can deliver an endless sequence of nudity and bar fights?

If we return to the paperback example, however, we might find a small sliver of hope. Ultimately, the explosion of these cheaper, often lower-quality books didn’t lead to the elimination of more serious titles. In fact, the opposite happened. Vastly more hardcover titles are published today than they were before the Pocket Books revolution began.

A closer look reveals that by vastly increasing the market for the published word, paperbacks also vastly increased the opportunities to make a living writing serious books (which, for the sake of this discussion, I’ll define as books that require at least a year to write and are published in hardcover). There was, to be sure, a lot of trash put out during the heyday of the paperback, but this reconfigured publishing model also generated a lucrative secondary market for more traditional writers.

Stephen King, for example, sold the hardcover rights to his first novel, Carrie, for around $2,500 in 1973 ($18,000 in today’s dollars). This was a nice bonus, but hardly enough to live on. The paperback rights for Carrie, by contrast, sold for $400,000 (almost $3,000,000 in today’s dollars), allowing King to quit his day job and become a full-time writer.

King wasn’t alone; other acclaimed authors, from Ursula K. Le Guin to Ray Bradbury, to Agatha Christie, also would have never risen to such prominence without the opportunities provided by the paperback world. As for Crichton, we know what happened next. The nine, mostly cheesy paperbacks, he wrote using pseudonyms, helped him polish his craft. His first hardcover book, The Andromeda Strain, was a massive bestseller and initiated the beginning of a career as one of the most influential writers of his generation.

As you know, I strongly dislike much of the current digital attention economy, and I believe that most people should be spending vastly less time engaging with these products. But in the spirit of trying to end 2025 on an optimistic note, I find some solace in the story of paperback books. Just because a certain type of low-quality media becomes immensely popular doesn’t necessarily mean that the deeper alternatives will suffer. Over one billion TikTok videos will be viewed today, and yet, you’re still here, reading a speculative essay about media economics. I don’t take that for granted.

The post On Paperbacks and TikTok appeared first on Cal Newport.

View the full article

The law requires these companies to identify and deactivate accounts of users under 16, and to prevent them from setting up new accounts in the future. Failure to comply can result in fines of up to $33 million.

Since it was proposed a year ago, the ban has drawn complaints from tech companies who argued that determining users’ ages is somehow beyond their engineers’ capabilities. There was also scattered pushback from civil liberties groups concerned about privacy and free speech.

But the government remained firm, ​stating​ it was committed to its goal of combating “design features that encourage [kids] to spend more time on screens, while also serving up content that can harm their health and wellbeing.”

It was hard for them to do anything else after a study they commissioned earlier this year revealed the following disturbing trends:

• 96% of children aged 10-15 in Australia use social media
• 7 out of 10 had been exposed to harmful content.
• More than half had been the victim of cyberbullying.
• 1 in 7 experienced grooming-type behavior.

The natural follow-up question for Americans is: Would such a ban be legally feasible in our country? (Putting aside, for now, the political appetite for such regulation, which is a different issue altogether.)

Last January, ​I investigated this question for The New Yorker​. For this piece, I interviewed Meg Jones, a colleague of mine at Georgetown’s ​Center for Digital Ethics​, who works at the intersection of technology and law. I asked her to explain how regulators decide when it’s appropriate to ban something that harms kids.

I recommend reading ​the full article​ to learn more about the legal issues at play. But for those who are looking with a wistful eye toward our friends down under, I’ll reproduce Jones’s prediction. “I think age verification is going to pass constitutional scrutiny this year, and we’re going to see a wave of state laws restricting social media for kids,” she told me. “Or maybe that is just my wishful thinking.” I’m wishing, too.

The post Australia Just Kicked Kids Off Social Media. (Is the U.S. Next?) appeared first on Cal Newport.

View the full article

“A spooky convergence is happening in media,” he begins. “Everything that is not already television is turning into television.”

Thompson then gives three examples of what he means:

1. Social Media is moving from offering connection to streaming videos (in ​court documents​ from this summer, Meta admitted that only 7% of activity on their Instagram platform involves users following people they know).

2. Podcasts are migrating inexorably toward video format.

3. Even AI is shifting toward visual media with the launch of new products like ​OpenAI’s Sora​ and Meta’s Vibes.

Television, of course, can mean many things. “When I say ‘everything is turning into television,’” Thompson clarifies, “what I mean is that disparate forms of media and entertainment are converging on one thing: the continuous flow of episodic video.”

While I might disagree with the premise that all media is turning into television (the global book and movie industries continue to chug along well enough in their traditional formats), I think Thompson is pointing out a real and very important trend that’s particularly pronounced in the world of internet media. (I’ve been arguing this point for a while now; e.g., the first 15 minutes of my most recent ​appearance​ on the Tim Ferriss Show, when I tried to convince Tim that video was inevitably going to devour audio podcasts.)

This leaves us with a significant question: why are these media devolving into glorified TV? In other words, why has the internet, which once held promise as the collaborative, intellectually stimulating alternative to the boob tube, increasingly churned out a stream of TikTok clones?

I’ve come to believe that the answer is less about technological determinism than it is economic determinism. The video entertainment sector is shockingly lucrative, with the combined online video and traditional TV markets ​projected​ to reach $1 trillion in annual revenue by 2030. The reason why internet media are becoming more like TV, then, is because that’s where the money is!

The broadband and wireless internet infrastructure we built through public/private partnerships over the last two decades turned out to be the perfect foundation on which a small number of tech companies could build competitors to the existing television and video market – to grab their share of that trillion-dollar sector. You can’t blame Meta, or Google, or even OpenAI for chasing that market. It was too big to ignore.

But here’s the silver lining: Once we realize that these companies’ apps are essentially glorified TV, we should feel more comfortable ignoring them. There was a time when platforms like Facebook and Twitter wanted to convince you that they were part of a new social fabric; a fundamental technology that responsible citizens couldn’t ignore. Not any more. If they’re just TV, then we can respond the way we always have: by simply turning off the proverbial set.

The post Why is the Internet Becoming TV? appeared first on Cal Newport.

View the full article

Here’s how Grann responded:

“Coming up with the right idea is the hardest part. First, you try to find a story that grips you and has subjects that are fascinating. Then, you ask: Are there underlying materials to tell that story?… The third level of interrogation is: Does the story have another dimension, richer themes, or trap doors that lead you places?”

He later adds:

“I spend a preliminary period ruthlessly interrogating ideas as I come across them, even though it’s time-consuming and a bit frustrating. I don’t want to wake up two years into a book project saying, ‘This isn’t going anywhere.’”

These quotes caught my attention because their relevance extends beyond the craft of writing and to the broader concern of cultivating depth in a world increasingly mired in digitally-enhanced shallowness.

In life, the types of deep projects that free us from these shallows–whether by transforming our career into something remarkable or making our personal lives richer–require a massive investment of time and effort. This includes:

• Diligence – the willingness to stick with the pursuit for months or even years; and
• Deliberateness – the willingness to do the specific types of activities that actually matter for making progress, even if they’re hard, and there are other activities you’d rather do instead.

Given these demands, it’s common to either lose interest in these projects once they get going or to be so intimidated by the path ahead that you never get started in the first place.

Grann’s advice helps with both issues. By raising the bar for considering a deep project–relentlessly examining, researching, and studying the reality of a pursuit before finally deeming it worthy–you’ll naturally end up giving serious consideration to fewer ideas. And those that do make it through this gauntlet will be so compelling that you’re much more likely to get started and stick with them.

This pre-commitment vetting is often a missing piece when discussing grand goals. Online “hustle culture” voices often emphasize activity for its own sake: Get started! Delay is for the weak! Craftsmen like Grann, on the other hand, understand that fundamental to the art of deep accomplishment is the patient search for the right subject.

The post David Grann and the Deep Life appeared first on Cal Newport.

View the full article

Somers then talks to experts who study how modern large language models operate, and notes that the mechanics of LLMs’ next-token prediction resemble this existing definition of thinking. Somers is careful to constrain his conclusions, but still finds cause for excitement:

“I do not believe that ChatGPT has an inner life, and yet it seems to know what it’s talking about. Understanding – having a grasp of what’s going on – is an underappreciated kind of thinking.”

Compare this thoughtful and illuminating discussion to another recent description of AI, delivered by biologist Bret Weinstein on an episode of Joe Rogan’s podcast.

Weinstein starts by (correctly) noting that the way a language model learns the meaning of words through exposure to text is analogous to how a baby picks up parts of language by listening to conversations.

But he then builds on this analogy to confidently present a dramatic description of how these models operate:

“It is running little experiments and it is discovering what it should say if it wants certain things to happen, etc. That’s an LLM. At some point, we know that that baby becomes a conscious creature. We don’t know when that is. We don’t even know precisely what we mean. But that is our relationship to the AI. Is the AI conscious? I don’t know. If it’s not now, it will be, and we won’t know when that happens, right? We don’t have a good test.”

This description conflates and confuses many realities about how language models actually function. The most obvious is that once trained, language models are static; they describe a fixed sequence of transformers and feed-forward neural networks. Every word of every response that ChatGPT produces is generated by the same unchanging network.

Contrary to what Weinstein implies, a deployed language model cannot run “little experiments,” or “want” things to happen, or have any notion of an outcome being desirable or not. It doesn’t plot or plan or learn. It has no spontaneous or ongoing computation, and no updatable model of its world – all of which implies it certainly cannot be considered conscious.

As James Somers argues, these fixed networks can still encode an impressive amount of understanding and knowledge that is applied when generating their output, but the computation that accesses this information is nothing like the self-referential, motivated, sustained internal voices that humans often associate with cognition.

(Indeed, Somers specifically points out that our common conceptualization of thinking as “something conscious, like a Joycean inner monologue or the flow of sense memories in a Proustian daydream” has confused our attempts to understand artificial cognition, which operates nothing like this.)

~~~

I mention these two examples because when we talk about AI, they present two differing styles.

In Somers’s thoughtful article, we experience a fundamentally modern approach. He looks inside the proverbial black box to understand the actual mechanisms within LLMs that create the behavior he observed. He then uses this understanding to draw interesting conclusions about the technology.

Weinstein’s approach, by contrast, is fundamentally pre-modern in the sense that he never attempts to open the box and ask how the model actually works. He instead observed its behavior (it’s fluent with language), crafted a story to explain this behavior (maybe language models operate like a child’s mind), and then extrapolated conclusions from his story (children eventually become autonomous and conscious beings, therefore language models will too).

This is not unlike how pre-modern man would tell stories to describe natural phenomena, and then react to the implication of their tales; e.g., lightning comes from the Gods, so we need to make regular sacrifices to keep the Gods from striking us with a bolt from the heavens.

Language model-based AI is an impressive technology that is accompanied by implications and risks that will require cool-headed responses. All of this is too important for pre-modern thinking. When it comes to AI, it’s time to start our most serious conversations by thinking inside the box.

The post When it Comes to AI: Think Inside the Box appeared first on Cal Newport.

View the full article

Here’s the problem: this is also the address where my readers send me interesting notes about my essays, or point me toward articles or books they think I might like. I want to read these messages, but they’re often hidden beneath unruly piles of digital garbage.

So, I decided to see if AI could solve my problem.

The tool I chose was called ​Cora​, as it was among the more aggressive options available. Its goal is to reduce your inbox to messages that actually require your response, summarizing everything else in a briefing that it delivers twice a day.

Cora’s website notes that, on average, ninety percent of our emails don’t require a reply, “so then why do we have to read them one by one in the order they came in?” Elsewhere, it promises: “Give Cora your Inbox. Take back your life.”

This all sounded good to me. I activated Cora and let it loose.

~~~

I detail the story of my experience with Cora in my latest article for The New Yorker, which is titled ​“Why Can’t A.I. Manage My E-Mail?”​, and was published last week.

Ultimately, the tool did a good job. This inbox has indeed been reduced to a much smaller collection of messages that almost all actually interest me. The AI is sometimes overzealous and filters some messages that it should have left behind, but I can find those in the daily briefings, and nothing that arrives here is urgent business, so the stakes are low.

The bigger question I ask in this article, however, is whether AI will soon be able to go beyond filtering messages to answering them on our behalf, automating the task of email altogether. This would be a big deal:

I’ve come to believe that the seemingly humble task of checking e-mail—that unremarkable, quotidian backbeat to which digital office culture marches—is something more profound. In 1950, Alan Turing argued in a seminal paper that the question “Can machines think?” can be answered with a so-called imitation game, in which a computer tries to trick an interrogator into believing it’s human. If the machine succeeds, Turing argued, we can consider it to be truly intelligent. Seventy-five years later, the fluency of chatbots makes the original imitation game seem less formidable. Yet no machine has yet conquered the inbox game. When you look closer at what actually goes into this Sisyphean chore, an intriguing thought emerges: What if solving e-mail is the Turing test we need now?

Cora, as it turns out, cannot solve the Inbox Game – it can organize your messages, but not handle them on your behalf. Neither can any other tool I surveyed, from SuperHuman to SaneBox. As I go on to explain in my article, this is not for lack of trying: there are key technical obstacles that make answering emails something AI tools aren’t yet close to solving.

I encourage you to ​read my full article​ for the entire computer science argument. But I want to emphasize here the conclusion I reached: even with their current constraints, which limit AI-based tools mainly to filtering and summarizing messages, there’s still much room for them to evolve into increasingly interesting and useful configurations.

In my article, for example, I watched a demo of an experimental AI tool that transforms the contents of your inbox into a narrative “intelligence briefing.” You then tell the tool in natural language what you want it to do – “tell Mary to send me a copy of that report and I’ll take a look” – and it writes and sends messages on your behalf. The possibilities here are intriguing!

Here’s how I ended my piece:

“Although A.I. e-mail tools will probably remain constrained…they can still have a profound impact on our relationship with a fundamental communication technology. …Recently, I returned from a four-day trip and opened my Cora-managed inbox. I found only twenty-four new e-mails waiting for my attention, every one of them relevant. I was still thrilled by this novel cleanliness. Soon, a new thought, tinged with some unease, crept in: This is great—but how could we make it better? I’m impatient for what comes next.”

This is the type of AI that interests me. Not super-charged chatbot oracles, devouring gigawatts of energy to promise me wise answers to any conceivable query, or the long-promised agents that can automate my tasks completely. But instead, practical improvements to chores that have long been a source of anxiety and annoyance.

I don’t need HAL 9000; an orderly inbox is enough for now.

The post Why Can’t AI Empty My Inbox? appeared first on Cal Newport.

View the full article

I remember what happened next because ​I wrote an essay​ about the experience. The tale starts slow:

“It was early, and the fog was just starting its march down the Berkeley hills. I eventually wandered into an eucalyptus grove. Once there, I sipped my coffee and thought.”

I eventually come across an interesting new technique to circumvent a key mathematical obstacle thrown up by The Beast. But this hard-won progress soon presented a new issue:

“I realized… that there’s a limit to the depth you can reach when keeping an idea only in your mind. Looking to get the most out of my new insights, and inspired by my recent commitment to the textbook method, I trekked over to a nearby CVS and bought a 6×9 stenographer’s notebook…I then forced myself to write out my thoughts more formally. This combination of pen and paper notes with the exotic context in which I was working ushered in new layers of understanding.”

I even included a nostalgically low-resolution photo of these notes:

More than a decade later, I can’t remember exactly which academic paper I was working on in that eucalyptus grove, but based on some clues from the photo above, I’m pretty sure it was this ​one​, which was published the following year and received a solid 65 citations.

I revisited this essay on my ​podcast​ this week. The activity it captured seemed a strong rebuke to the current vision of a fast-paced, digitized, AI-dominated workplace that Silicon Valley keeps insisting we must all embrace.

There’s a deeply human satisfaction to retreating to an exotic location and wrestling with your own mind, scratching a record of your battle on paper. The innovations and insights produced by this long thinking are deeper and more subversive than the artificially cheery bullet points of a chatbot.

The problem facing knowledge work in our current moment is not that we’re lacking sufficiently powerful technologies. It’s instead that we’re already distracted by so many digital tools that there’s no time left to really open the throttle on our brains.

And this is a shame.

Few satisfactions are more uniquely human than the slow extraction of new understanding, illuminated through the steady attention of your mind’s eye.

So, grab a notebook and head somewhere scenic to work on a hard problem. Give yourself enough time, and the enthusiastic clamor about a world of AI agents and super-charged productivity will dissipate to a quiet hum.

The post Forget Chatbots. You Need a Notebook. appeared first on Cal Newport.

View the full article