Do Hard Things (Featuring: Jerry Seinfeld)

The episode begins with Cal Newport recounting a story from the Acquired podcast’s episode on the business of Seinfeld. In an interview, a Harvard Business Review interviewer asked Jerry Seinfeld why he ended the show while it was so successful, rather than finding a more efficient and sustainable way to produce it. The interviewer jokingly asked what McKinsey, the large consulting firm, might have recommended if it had examined the show’s workflow. Seinfeld replied, “Who’s McKinsey?” When told that McKinsey was a big consulting firm, he asked, “Are they funny?” After the interviewer answered no, Seinfeld said, “Then I don’t need them.” When asked whether he had no interest in efficiency, Seinfeld responded, “No, no. If you’re efficient, you’re doing it the wrong way. The right way is the hard way.” Newport uses this exchange as the episode’s central idea: in creative and cognitive work, the difficult process may be the part that creates the value.

Newport argues that workplace technologies are not merely optional tools that help people perform an unchanged job. They can also reshape people’s implicit understanding of what work is. He points to his 2021 book, A World Without Email, which examines how email changed office work beyond simply replacing fax machines and voicemail. Email was initially promoted as a pragmatic improvement: instead of sending paper through a fax machine or dialing into voicemail and listening to messages sequentially, workers could send a message that would wait in the recipient’s inbox. But Newport says the deeper effect was to change the rhythm of work. Before email, an employee might work individually toward a concrete result and occasionally collaborate in a meeting. After email, work increasingly became a continuous back-and-forth exchange—asking for information, reviewing drafts, and responding to requests as work unfolded. As a result, being busy became associated less with completing substantial outputs and more with participating in ongoing communications. Newport says this created a damaging cycle: workers had to service their inboxes throughout the day, constantly switch attention, and remain involved in multiple asynchronous conversations. Slack intensified the same “hyperactive hive mind” model without, in his view, clearly increasing the value or quality of knowledge-work output.

Newport then applies this framework to workplace AI in 2026, especially AI agents modeled on the systems used by computer programmers. He argues that these tools are implicitly pushing a new definition of knowledge work: a cognitive version of the assembly line. Instead of manufacturing physical products such as cars or washing machines, the cognitive assembly line produces and processes tasks. Under this model, work is understood as a series of discrete actions: updating a slide deck, gathering numbers, inserting bullet points into an email, collecting information, drafting a report, or sending a message. AI agents, large language models, chatbots, and systems that orchestrate multiple agents can supposedly speed up each step. More tasks completed per day then appears to equal greater productivity. Newport warns that this model may be fundamentally wrong for many knowledge workers. The worker becomes a supervisor of AI outputs, checking, correcting, and coordinating small tasks without necessarily doing meaningful human work. He compares this to being a manager whose only interactions are with the outputs of agents and chatbots rather than with other people. The result, he argues, could be another miserable definition of work, similar to the exhaustion created by email and Slack.

Newport’s proposed defense against technology-driven changes is to develop a strong personal understanding of what matters in one’s work. If someone knows what “doing my work well” actually means, a new tool is less likely to redefine the job around whatever the tool can measure or accelerate. This is where he returns to Seinfeld’s statement that efficiency is the wrong goal and that “the right way is the hard way.” For Seinfeld, Newport argues, the valuable work was the serious application of cognition required to produce a strong script. That could involve combining things learned over time, using human instincts, generating creative possibilities, iterating through different configurations, editing, and curating the result. The outcome might be the equivalent of a tight Seinfeld script—or, in another profession, an advertising pitch that Kodak would love. That process is difficult, and whether it takes 19 days or 21 days is much less important than whether it produces something genuinely valuable. Newport says that this model focuses attention on a small number of high-value outputs rather than on marginal gains in task speed. If an email system makes messages move faster but does not help Seinfeld write a better script, it is not central to his work. Likewise, if AI agents reformat scripts, locate jokes, or update status spreadsheets but do not improve the script itself, their speed may not matter and could even distract from the difficult work that counts.

Newport acknowledges that his Seinfeld analogy does not apply equally to every occupation. Some jobs may be well suited to constant communication or to AI agents coordinating tasks. But he believes more cognitive jobs than people realize have a core activity that resembles Seinfeld’s work: they depend on doing a particular difficult thing exceptionally well, and the quality of that thing drives results for the organization and its customers. The recommendation is not to reject technology or become a Luddite. Instead, workers should keep first things first, deploy tools selectively, and be reluctant to adopt systems merely because they are new or efficient. Newport says that working on the technological cutting edge often means spending substantial time figuring out the technology itself. It may be better to wait until the cutting edge has passed and then adopt tools in strategic ways with a better cost-benefit ratio. He jokes that Jerry Seinfeld probably does use email, but he was not trying to build the world’s first email servers while writing early Seinfeld scripts. In the same way, an AI tool can be useful without becoming the central metaphor for an entire job. Newport closes by arguing that many cognitive workers should seek out the “hard” part of their work rather than automatically trying to eliminate it. He notes that Seinfeld’s show has reportedly generated somewhere between $4 billion and $5 billion in revenue since its conception, suggesting that Seinfeld’s emphasis on doing the difficult work was not obviously misguided.