Smiling Exhaustion: Inside the Tech Worker Sentiment Survey as AI's Honeymoon Fades

Lenny Rachitsky and Noam Segal, his former research partner at Airbnb, discuss the second Tech Worker Sentiment Survey. This year it reached about 6,000 people across product, engineering, design, research, marketing and other roles. Last year's edition was titled "burnt out but optimistic", and Segal says he knows of no other survey of this kind at this scale.

The headline finding is how bifurcated the workforce is. Segal says AI shapes how people feel about work more than their role, company, company size or seniority. Asked how AI has shifted their professional identity, only 3% said it hasn't, 50% feel amplified, 27% feel their role is being redefined without clarity, 14% feel destabilized and 5% feel diminished.

Segal puts the size of the effect in perspective by using effect sizes instead of statistical significance, which can be reached with any large sample. Last year's biggest effects were the quality of your manager and the unusual happiness of founders. The AI identity effect is about three times as large as either, and it lines up neatly with career optimism, burnout, layoff worry and willingness to recommend one's role.

To describe the emotional picture, the team asked respondents which emotions they feel at work; on average people picked five. From those answers they derived four archetypes. The energized (41%) say product has become fun again and feel like they are in a tech amusement park. The conflicted (35%) are having the most fun of their careers while feeling the most uncertainty, unsure whether they are building the tools that end their own careers.

The disoriented feel their role keeps shifting. One respondent likened the moment to farmers on the cusp of the industrial revolution. The resentful (12%) feel pressured and checked out, with comments like being forced to use AI or lose their job, then watching people lose their jobs anyway.

Rachitsky and Segal both stress empathy across these groups. Segal worries that too much attention goes to mastering new models and techniques and too little to relationships at work. He asks energized people to spend some of that energy supporting colleagues who are struggling.

Segal says he hoped better models would let people work less hard, but the data show the opposite. Burnout above a moderate level climbed from 44.7% in 2025 to 54.7% in 2026, and optimism about the future of one's role fell from 54.8% to 48.7%. Rachitsky calls the ten-point burnout jump alarming.

Segal contrasts this with the classic theory that stagnation causes burnout, citing Rachitsky's earlier guest from Ramp, who said his worst burnout came when velocity was lowest. Now people are shipping faster than ever, going from a couple of PRs a day to dozens, and taking on more prototypes, PRDs, campaigns and agents. They aren't working less, just taking on more, and burnout rises with it.

There is a glimmer of hope in the data: enjoyment of work stays high. People say AI lets them step out of their classic swim lanes, such as a PM tapping their inner designer or a marketer their inner engineer. They can also build things that seemed impossible not long ago, at work or on the side.

Fear of layoffs sits alongside the enjoyment. About 72% of respondents are worried to some degree about being laid off and 41.2% are at least moderately worried, and some feel they are cutting the branch they sit on. Segal calls the technology addictive, since it is hard to put down, tiring to keep learning and exciting precisely because it is so powerful.

Rachitsky, a self-described NPS hater, and Segal still asked respondents whether they would recommend their role to a friend or relative entering tech. On the scale, zero is neutral, and no function reached it. Founders, sales and go-to-market, product, operations and engineering all came out as detractors, while designers and researchers were the least likely to recommend their roles.

Willingness to recommend also falls with seniority, so executives, VPs and directors are more likely to recommend their roles than individual contributors. Segal suggests senior leaders benefit from AI digesting streams of information, while ICs scramble to build micro-products with duplicated effort and a heavy maintenance burden. He compares it to pushing the pedal down in neutral gear.

Segal frames the mood with a ladder metaphor drawn from Cognition CEO Scott Wu's description of Devin, which climbed from high school CS student to intern to junior to senior engineer. The higher you already are on the ladder, the safer you feel. Everyone below feels the rungs being pulled from under their feet, so it makes sense not to recommend climbing.

Segal, a researcher by training, finds the research result painful. He argues AI has lowered the floor but not raised the ceiling, and that research and design matter more as building gets cheap. Both hosts caution that this measures feelings, not reality, and doesn't mean these roles are going away.

Most people still say they enjoy their current role. Segal reads the gap as a difference between present experience and extrapolation: people are doing okay today, but what they expect in one, three or five years is much more negative.

When asked how much better AI makes them at their jobs, 97.2% said it does and nearly half said very much or extremely. But Segal says the follow-ups revealed that better means something different from what he assumed. People mean they can produce more faster, whether PRs, PRDs, research projects or prototypes, but not that the output is of higher quality.

The second concern is the cost to thinking and judgment. Respondents say things like "my brain is rotting, my work feels worse." Segal ties this to cognitive rot, where people accept the model's first output without applying their own judgment and let their involvement and agency collapse.

Segal suggests one explanation is that the honeymoon with AI is over. Models are the worst they will ever be, but they still have a way to go. Referencing Simon Willison's warnings about skill atrophy, he says the data suggest we are losing that fight a little and need deliberate practice, thinking and judgment. Rachitsky adds that people take the easy button when it's there, and wonders how juniors will learn to code or write a PRD from scratch.

Segal adds a psychological angle: every time you solve a problem yourself you raise your baseline of self-efficacy, and every time you hand it to a model you lower it. He closes with a line from the report, that the productivity gains are real but the quality of the work and the sharpness of the person producing it are taking a hit. Rachitsky jokes that humans now need evals.

The dominant public narrative is the fear of replacement, but respondents ranked losing their job to AI second to last among their fears. At the top was the expectation to do more for the same pay. Second was the sense that the pace is becoming unsustainable, both the pace of work output and the pace of technological change.

Segal describes a downward spiral. Learning each new model, framework and method eats into the focused time needed to hit ever-rising output expectations. Rachitsky reads another line from the report: the speed AI unlocked got plowed straight back into expectations, and every gain becomes the new baseline.

The emotions people report are a wide mix, with curiosity and excitement at the top and overwhelm, conflict, relief, tiredness, anxiety and hope also appearing. Segal cites Nikhil Singal's phrase "smiling exhaustion" as the best summary. People feel reborn as builders but have no off switch, and the rules keep rewriting themselves.

Segal pushes back on black-and-white thinking about hypers and doomers. He says everyone holds both in different amounts, and that it is natural to be excited and overwhelmed, relieved and tired, even resentful, all at once.

As in last year's survey, designers and researchers are the most negative group. They are the most likely to feel destabilized or diminished, to feel tired, overwhelmed and anxious, to worry about losing their jobs and to advise against entering their roles. Segal, who belongs to both communities, argues the industry needs them, pointing to the importance of taste, craft and quality as voiced by guests like Katie Dill and Jenny Wen.

Rachitsky says he expected this to be design's moment, given how generic AI-made products look. Segal notes that some of AI's limits in design may be intractable, and that novel creative experiences are still beyond the technology. Data analysts are actually the most worried about losing their jobs, which Segal says reflects how people feel rather than a one-to-one measure of AI capability.

The happiest people are again founders and people at smaller companies. Founders are the most optimistic, at 71%, enjoy their jobs most and have the lowest burnout and layoff worry, although 47% are still at least moderately burnt out. Segal adds a selection-bias caveat that these are founders whose companies are still running.

Company size shows a strikingly linear pattern. As employer size grows from a 1-10 person startup to an enterprise of thousands, optimism falls and burnout and layoff worry rise, and willingness to recommend one's role falls. Segal still calls even the lowest burnout levels worrisome and says the vibe is very different from when he and Rachitsky worked at Airbnb a decade ago.

For the second year, manager effectiveness stands out. The more effective your manager, the less burnt out you are and the more you enjoy your job. People with an extremely effective manager report around 65% higher job enjoyment and dramatically lower burnout.

The problem is supply. Only about 25% of respondents rate their manager as highly effective, and 36% rate them as ineffective, little changed from last year. Segal worries that the era of the great flattening and founder mode, with more direct reports per manager, undercuts the one lever that matters most. Managers are also the people who shield employees from the AI squeeze.

The worst-rated managers cluster in data and analytics and in design. Segal's guess is that managers are struggling with the same feelings as everyone else and passing some of that on to their teams. He urges executives to invest in manager training, and notes how hard it is to build great managers only to see them poached by the labs.

Respondents were asked to describe the state of the tech industry in their own words, and the resulting word cloud is full of change, chaos, speed, hype, instability, bubble and opportunity. One quote was "We're in the second inning of a massive shift. No one knows how it will end, but all you can do is keep taking at bats." Another called tech manic, with half out of touch and clinging to the bandwagon and the other half exhausted by the first.

Sentiment analysis of the words came out at 37% positive, 37% negative and about 26% neutral. Segal describes the split as everyone seeing the same thing, with half finding it thrilling and half terrifying. Rachitsky muses that we are living through history, and Segal declines to prophesy about whether this is the most normal things will ever be.

For his closing thought, Segal says that having the most advanced AI and the best models won't determine whether an organization succeeds. Underneath it all are people living through the biggest shift of their lives, excited and exhausted and hopeful and scared at once. He tells listeners to take care of themselves, touch grass and reflect on what they want from their careers.

For employees, the first piece of advice comes from the amplified and energized respondents, who went deep on specific tasks instead of trying to be a generalist who does everything. Segal warns that leaning too hard into the generalist narrative and ignoring the core of your role leads to severe burnout. He also urges people to watch for being overworked for the same pay and to take the burnout test linked in the article.

Other advice is to talk to your manager when scope grows with no change in compensation and to protect and invest in that relationship. He also suggests considering a smaller company or starting your own, and, for early-career people, seeking mentors and teams willing to invest in developing you.

For leaders, Segal says to invest in managers, manage the squeeze by setting sustainable expectations and keep the bottom rung of the ladder alive for early-career hires. Those hires are often the most AI-native people around. Leaders should also pay attention to roles like design and research that are feeling more negative, since AI is lifting some people and destabilizing others.

Segal also flags an emerging AI guilt, especially among early-career people and in product marketing and data and analytics, who feel that using AI is a kind of cheating. He ties it to impostor syndrome, with success attributed to a technology rather than to other people. His message is that AI is the worst it will ever be and is only getting better, so there is no reason to feel guilty about using it.