The stronger the AI, the more tired people become, and 'anxiety' becomes the norm for companies and employees Article: Xu Chao Source: Wall Street Insight AI programming tools promise to liberate engineers, but reality has given rise to a new round of efficiency anxiety. As the capabilities of AI programming agents such as Anthropic's Claude Code and OpenAI's Codex continue to soar, technology companies are caught in a top-down 'productivity obsession'. The executives personally went off to write the code, and the employees were asked to refresh the frequency of interaction with AI, while the overtime hours did not decrease but increased. AI should have been a labor-saving tool, but it has become a new source of stress in many workplaces. Survey data reveals a clear cognitive gap: According to a survey by consulting firm Section, over 40% of C-level executives believe that AI tools save them at least 8 hours per week, while 67% of non management employees say that AI helps them save less than two hours of time, or even no help at all. A continuous study conducted by the University of California, Berkeley on a 200 person organization found that even though employees have transferred a large amount of work to AI, their actual working hours are still increasing. The spread of this anxiety has structural reasons. When chief technology officers type code into AI at 5am and CEOs measure team effort by billing amounts, the industry's imagination of "efficiency" has been redefined - and the cost of this redefinition is being borne by ordinary employees. Executives enter to write code, and efficiency anxiety spreads from top to bottom. The term 'Vibe coding' initially carries a lazy expectation. Former OpenAI researcher Andrej Karpathy introduced this concept to the public eye in February 2025, describing a new programming mode where engineers can complete development by simply chatting with AI - 'completely immersed in the atmosphere'. However, a year later, the atmosphere had already changed. Intuit Chief Technology Officer Alex Balazs described his recent daily routine: his wife came downstairs at 8am and found him working for several hours. She asked me how long I had been up, and I said I had been writing code since 5am. To be precise, he was guiding the AI agent to write code for him, which he said allowed him to delve deeper into the underlying code that he hadn't touched for many years. This type of executive behavior is transmitting downward pressure. OpenAI President Greg Brockman recently posted on X, saying, "Every moment your agent isn't running, it feels like a waste of opportunity." This statement accurately triggered the already prevalent workaholic culture in the technology industry. Alex Salazar, co-founder and CEO of AI startup Arcade.dev, is more direct. He regularly checks the company's Claude Code bills - the bill amount is directly linked to the frequency of engineers using the tools - and specifically criticizes employees who "don't spend enough": "I would say, 'You're not putting in enough effort.'" He said that after the first such "faith meeting," the company's AI programming tool bill skyrocketed tenfold, and he sees this expense as a sign of progress. Employees are being quantitatively managed, and 'AI fatigue' is quietly spreading in this atmosphere, and the way employees are assessed is also quietly changing. DocuSketch is a software company focused on property restoration business, and its Vice President of Product, Andrew Wirick, stated that the company now tracks the "number of interactions" engineers have with AI programming tools on a daily basis, and the higher the default number, the stronger the team productivity. Claude Code will also generate weekly reports for each engineer, listing all patterns in which they have fallen into ineffective loops with AI and providing improvement suggestions. Wirick himself admitted that he has developed a certain sense of addiction. I feel like I have to complete a few more interactions every day, and before going to bed, I'm still thinking about how to do a few more. "He attributes this state to his" epiphany experience "when he tried out Anthropic's latest model Opus 4.5 in November last year - at that time, he handed over a functional prototype task that usually needs to be handed over to engineers to the model, and after 20 minutes, he saw the model autonomously disassemble and implement the task," feeling like the brain was rebooted. This mentality of accelerating everyone is eroding the boundaries between work and life. Berkeley's research found that even though a large number of tasks have been taken over by AI, people's working hours have not shortened. Some engineers have also begun to openly admit that they are experiencing "AI fatigue" - constantly worrying about missing the next breakthrough, which seems to be only one prompt word away. The cognitive gap between executives and employees is widening, and the enthusiasm of executives largely comes from the freshness created by themselves. Salazar admits that personally using AI to build prototypes is more "productive and immediate" than handling authorization and decision-making on a daily basis. He recently even directly responded to a service request from an important financial client and built a demo application from scratch. At Intuit, product managers and designers are now encouraged to use "Vibe coding" to build functional prototypes on their own in QuickBooks. Balazs said, "At least now, product managers can hold a specific thing and say to engineers, 'I want something like this.' However, survey data from Section Consulting shows that this cognitive gap is quite significant. There is a huge gap between executives' perception of the benefits of AI and the experience of grassroots employees. Salazar believes that this is partly due to employees bearing higher transformation costs when adapting to new tools: "They are implicitly asked to find time to explore and experiment, but their daily work expectations are not adjusted accordingly to free up this space." The hidden concern of job security also exists. Salazar admitted that he had originally planned to switch to a third-party network service provider, but now the marketing team is able to update the company's website using AI tools on their own, and this outsourcing expenditure has been cut as a result. The phenomenon of "task expansion" and false prosperity, the other side of the efficiency myth. Researchers at Berkeley have named this phenomenon "task expansion": when non-technical colleagues start using AI to generate code, engineers have to spend time cleaning up these semi-finished products, which actually increases their workload. Intuit's Balazs admits that this is reshaping the previously clearly defined division of labor, leading to more and more roles becoming 'mixed' and making existing collaborative relationships more complex. The deeper question is whether this construction boom is creating something valuable or just creating more things? Analysts point out that if this AI driven productivity obsession is not restrained, it may lead to a large number of "busyware" (wasted software) emerging - small changes to websites that no one cares about, customized dashboards with only one user, and prototype projects abandoned by marketing managers, all of which will ultimately be implemented by engineers. Each item seems to have its own reasons at present, but most will eventually fall into the garbage bin of discarded code. According to Balazs from Intuit, the productivity of the company's engineers has increased by approximately 30% in terms of code production and delivery speed. But in this increasingly disposable future of code, the real efficiency dividend may lie in the answer to another question: which things should never be built at all.