Does a Dream Die When AI Fulfills It? Checking "Instant Dream Death," "Dream Deflation," the "Obsession Gap," "Passion Capitalism," and "Life Prototypes" Against the Academic Literature

Five claims in one post

On October 10, 2026, paji_a posted on X roughly the following (paji_a 2026a). The greatest tragedy of the AI era is not losing one’s job but having a dream pursued for 20 years come true in 20 minutes, which the author calls the “instant death of dreams” (yume no sokushi). Once the “three great excuses” (no talent, no time, no skills) are stripped away, many people may find that they fulfilled their dream and did not particularly enjoy it. Being able to make things will no longer be rare (“dream deflation”), and the more AI narrows the “ability gap,” the more the “obsession gap” will be exposed. What one can be abnormally absorbed in, rather than education, career history, or credentials, will become a person’s competitive edge (“passion capitalism”). Beyond this lies an “age of life prototypes,” in which anyone can prototype their own life again and again.

The post states that professional quality and commercial success are a separate matter. The author also says that each of the five terms was coined on his own, so none of them is an academic term. Even so, each term contains claims that can be tested. For example, instant dream death claims both that the time needed to make things shrinks greatly and that satisfaction after fulfillment is low. The obsession gap claims both that AI narrows ability gaps and that, as a result, differences in obsession acquire value.

This note divides the five claims into testable propositions and, for each proposition, sets the studies that support it against the studies that do not. The answer that remains is that, of the five, dream deflation is the best supported by research, and only with respect to the price of making things and the value of what made things signal. For the other four, either a plausible mechanism exists without direct testing, or the research points in the opposite direction.

Scope and method

• Coverage: peer-reviewed articles, scholarly books, and research reports corresponding to the five claims. The topics are the psychology of goal attainment and anticipation; ownership and intrinsic motivation when making things with AI; self-handicapping; the economics of signaling; markets for creative goods under generative AI; experiments on generative AI and productivity; the psychology of grit and passion; the sociology of choosing work by passion; implicit theories of interest; provisional selves and life design; and changeable decisions and satisfaction. The follow-up post (paji_a 2026b), which describes an “age of automatically generated desire” in which AI anticipates and creates desires themselves, is outside the scope.
• Collection: 67 candidates were gathered from scholarly databases (OpenAlex, Crossref, arXiv, NBER), and their bibliographic records, retractions, and withdrawals were checked. Toner-Rodgers (2024) was excluded because arXiv administrators withdrew it, citing concerns about the validity of the data among other reasons. Sources for which neither full text nor abstract could be obtained (Markus and Nurius 1986, Berglas and Jones 1978, Snyder and Higgins 1988, Brickman et al. 1978, Carver and Scheier 1990, Sheldon and Elliot 1999) were excluded, as was Dell’Acqua et al. (2026), whose full text could not be obtained and whose abstract was short.
• Verification: for 39 of the 47 sources, the full text (published version, author version, working paper version, or preprint) was obtained, and numbers and claims were checked against the relevant passages. Seven sources were used only within the scope of their abstracts and are attributed in the text with “according to the abstract.” The full text of Cech (2021) could not be obtained, and it is described within the scope of the publisher’s description. For preprints and working papers, the existence of a peer-reviewed version was checked, and those without one are marked “not peer reviewed” in the text.
• Writing: for each claim, the note gives the proposition, the supporting studies, the conflicting studies, and a judgment, in that order.

Instant dream death: does 20 years become 20 minutes, and does fulfillment fail to satisfy?

How much has the time to make things shrunk?

Large time savings have been measured for short tasks with a fixed scope. In a preregistered experiment, 453 college-educated professionals were given occupation-specific writing tasks, and half of them were allowed to use ChatGPT; average time taken fell by 40%, and rated quality rose by 18% (Noy and Zhang 2023, abstract of the published version). In a task of implementing an HTTP server in JavaScript, the group with access to GitHub Copilot finished 55.8% faster (95% confidence interval 21–89%; Peng et al. 2023, not peer reviewed).

However, when experienced developers worked on real projects they knew well, the result was reversed. Sixteen developers, with an average of five years of prior experience on mature open-source projects, completed 246 tasks, each randomly assigned to allow or disallow AI (Becker et al. 2025, not peer reviewed). Before the study, the developers forecast that AI would reduce completion time by 24%, and afterward they estimated that it had reduced completion time by 20%, but in fact tasks took 19% longer.

What these studies measured is the time for tasks that take tens of minutes to a few hours. No study was found that measured how much a goal pursued over years, such as making a game or writing a novel, is shortened. The post itself states that professional quality and commercial success are a separate matter. “Twenty years in twenty minutes” is therefore best read as an extrapolation from the time savings on tasks.

Is satisfaction after fulfillment smaller than anticipated?

The idea that the time before fulfillment has value in itself has long-standing support. Loewenstein (1987) formalized the positive utility derived from anticipating future consumption, which he called savouring. In a survey in the same paper, respondents were willing to pay more for a kiss from the movie star of their choice delayed by three days than for an immediate kiss. According to the abstract of Van Boven and Ashworth (2007), across five experiments, people reported more intense emotions when anticipating events than when recalling them.

Emotions after an event are often smaller and shorter than predicted. Wilson and Gilbert (2005) called the tendency to overestimate the intensity and duration of emotional reactions to future events the impact bias, and named as one cause focalism, the tendency to underestimate how much other events will influence one’s thoughts and feelings. However, it cannot be said that happiness always returns to its earlier level after an event. According to the abstract of Diener, Lucas, and Scollon (2006), happiness set points are not neutral, differ between people, can change under some conditions, and individuals differ in how they adapt to events.

What was attained also mattered. Niemiec, Ryan, and Deci (2009) surveyed young adults one and two years after college. They reported that attaining intrinsic aspirations, such as personal growth, close relationships, and community involvement, was positively related to psychological health. In contrast, attaining extrinsic aspirations, such as money, fame, and image, was not, and was instead positively related to indicators of ill-being. The post asks whether what people wanted was the work, the act of making, or the expectation of a future self who would someday be great. In light of this study, the direction of satisfaction after fulfillment may depend on the answer to that question.

Does a thing feel less like one’s own when one has not made it?

The claim that the act of making has value in itself is supported by experiments. Norton, Mochon, and Ariely (2012) showed that people who assembled IKEA boxes, folded origami, or built Lego sets valued their own creations about as highly as those of experts, and named this the IKEA effect. The effect, however, was limited to cases in which people completed the task through their own labor, and it disappeared when they built and then destroyed their creations or failed to complete them. Inzlicht, Shenhav, and Olivola (2018) reviewed many studies showing that effort is both a cost people avoid and a source of value, in that the same outcome can feel more rewarding when more effort was applied.

Lower ownership has been reported for things made with generative AI. In two studies by Draxler et al. (2024; 30 and 96 participants), users did not consider themselves the owners or authors of AI-generated text, yet did not publicly declare AI authorship. In the same work, greater influence over the text was associated with a higher sense of ownership. In an essay-writing experiment with 54 participants, the group that used an LLM reported low ownership of their essays in interviews and had difficulty quoting from essays they had written minutes earlier (Kosmyna et al. 2025, not peer reviewed).

A decline in motivation after collaborating with AI has also been reported. In four online experiments by Wu et al. (2025; 3,562 participants in total), collaborating with generative AI on tasks such as drafting work emails or performance reviews raised immediate performance, but the gain did not persist when participants performed the next task alone. Moving from collaboration to solo work decreased intrinsic motivation and increased boredom.

What did the three great excuses protect?

Phenomena close to “no talent,” “no time,” and “no skills” have been studied as self-handicapping. The meta-analysis by Schwinger et al. (2022) defines academic self-handicapping as a strategy in which students who fear failure, instead of increasing their effort, procrastinate or withdraw effort so that potential failure can be attributed to these handicaps rather than to stable characteristics such as low intelligence. Across 159 studies (81,630 participants), habitual self-handicapping was strongly associated with conscientiousness (r = .40), fear of failure (r = .39), neuroticism (r = .38), and general self-esteem (r = .34).

Under this definition, the post’s three great excuses can work as a defense against attributing failure to one’s own ability. No study was found that measured whether people move to new excuses or face the judgment of their ability when AI removes “no skills” and “no time.”

The time spent living a dream can also be seen differently. In four experiments by Kappes and Oettingen (2011), inducing positive fantasies about an idealized future lowered energy as measured by physiological and behavioral indicators. Kuehn and Corrigan (2013) called un- or under-compensated work carried out in the present, often for experience or exposure, in the hope that future employment will follow, hope labor. The post says that while proving one’s talent took ten years, people could live those ten years as a dream. These two studies indicate that those ten years can also become a fantasy that saps energy or labor that goes unrewarded.

Judgment on instant dream death

The compression of time has been measured only for short tasks with a fixed scope, and the opposite result exists for experienced workers on real work. That fulfillment may not feel good is plausible given the research on anticipatory utility, the impact bias, and the attainment of extrinsic goals. In addition, because ownership of things made with AI is low and motivation for solo work declines after collaboration, a dream fulfilled with AI may bring less satisfaction than one fulfilled through one’s own work. However, no study was found that measured the satisfaction of people who fulfilled a long-held creative goal with AI, compared with their expectations before fulfillment. Because ownership varies with the degree of involvement (Draxler et al. 2024), satisfaction may depend less on whether AI was used than on how much of the content the person decided.

Dream deflation: has being able to make things stopped being rare?

Price and volume

Market data show that the volume of creative works has increased and prices have fallen. A study of China’s leading art outsourcing platform used the unexpected leak of an advanced image-generation model as a natural experiment and found that the average price of anime-style images fell by 64%, while order volume rose by 121% and total revenue rose by 56% (Zhang, Yuan, and Xiong 2026, checked in the arXiv version). The growth came mainly from low-priced personal orders, and incumbent creators retained the majority of the market. In more than four million artworks by more than 50,000 users, adopting text-to-image AI raised creative productivity by 25% and the likelihood of receiving a favorite per view by 50%, but the average novelty of the content declined (Zhou and Lee 2024).

For incumbent creators, the change appeared as a loss of attention. On a platform for anime- and manga-style artwork, after the launch of a text-to-image model, uploads by incumbent illustrators who did not adopt AI as a primary tool declined significantly, whereas comic artists were less affected (Kim, Jin, and Lee 2026, not peer reviewed). As one mechanism, the authors point to a loss of viewer attention to illustrations, measured by bookmarks.

Experiments and a meta-analysis also show that works become more similar. In an experiment on short stories, stories written with ideas from an LLM were rated as more creative, better written, and more enjoyable, but they were more similar to each other (Doshi and Hauser 2024). In a meta-analysis of 28 studies (8,214 participants), people working with generative AI outperformed people working alone in creative performance (g = 0.27), but the diversity of their ideas fell substantially (g = −0.86; Holzner, Maier, and Feuerriegel 2025, not peer reviewed).

Value as a signal

When being able to make things becomes common, what a made thing conveys about its maker also decreases. Spence (1973) wrote that a signal distinguishes applicants only if the cost of signaling is negatively correlated with productive capability. Otherwise, everyone invests in the signal in the same way, and the signal cannot distinguish them. Using data from a freelance labor platform, Galdin and Silbert (2025, not peer reviewed) showed that before LLMs, employers tended to pay more for workers whose applications were tailored to the posting, and that this tendency disappeared after LLMs arrived. In the counterfactual of their structural model, when written applications no longer signal ability, workers in the top quintile of the ability distribution are hired 19% less often and those in the bottom quintile 14% more often.

Effort originally served as a signal. Inzlicht et al. (2018) wrote that because effort is visible to both the actor and observers and is difficult to fake, it functions as a signal of dedication and intention. When effort can no longer be read from a work, this signal also weakens.

Where deflation stops

At the same time, a premium for human making remains. In an experiment that randomly assigned a “human-created” or “AI-created” label to paintings actually made by AI, paintings labeled as human-created were rated higher on all four criteria (liking, beauty, profundity, and worth), and the difference was moderated by the story and the effort perceived behind the work (Bellaiche et al. 2023). According to the abstract of Millet, Buehler, Du, and Kokkoris (2023), across four experiments (1,708 participants), the same artwork labeled as AI-made was perceived as less creative and induced less awe.

From the audience’s side, an increase in works also brings gains. Aguiar and Waldfogel (2018) used the tripling of new music releases between 2000 and 2008 brought about by digitization and estimated that, when the appeal of new products cannot be predicted in advance, more new products bring large benefits to consumers.

Judgment on dream deflation

The decline in the price of making things and in the value of what made things signal is supported by market data and experiments. However, in at least one market, total revenue increased along with volume (Zhang et al. 2026), and a premium remains for human making and visible effort (Bellaiche et al. 2023). What is deflating is the signal of being able to make things, while the value of a person having spent time making something remains.

The obsession gap: when ability gaps narrow, are differences in obsession exposed?

Did AI narrow ability gaps?

Narrowing has been reported mostly in work with a fixed scope. In a study of the staggered introduction of a generative AI conversational assistant to 5,172 customer support agents, issues resolved per hour rose by 15% on average (Brynjolfsson, Li, and Raymond 2025). For the least skilled quintile, resolutions rose by 36%, while the most skilled workers saw no productivity increase. In writing tasks, ChatGPT also reduced inequality between workers (Noy and Zhang 2023). In the short-story experiment, less creative writers benefited more (Doshi and Hauser 2024). In the programming experiment, developers with less experience benefited the most (Peng et al. 2023, not peer reviewed).

Results without narrowing also exist. In a field experiment that randomly gave Kenyan entrepreneurs access to a GPT-4 business assistant, the average effect on revenues and profits could not be distinguished from zero. However, the effect for initially low performers was about 0.25 standard deviations lower than for initially high performers; low performers did about 10% worse, whereas high performers may have benefited by more than 15% (Otis et al. 2024, not peer reviewed). The authors write that the difference arose not from the questions posed to the AI or the advice received, but from which pieces of advice entrepreneurs selected and implemented. The slowdown of experienced developers (Becker et al. 2025) also shows that AI does not raise everyone’s floor equally.

Across the labor market as a whole, large movements have not yet been measured. In Danish administrative data, effects on earnings and hours two years after the launch of ChatGPT were precise nulls that rule out effects larger than 2%, and what changed was the reorganization of tasks (Humlum and Vestergaard 2025, not peer reviewed).

What becomes visible when gaps narrow?

Even if ability gaps narrow, differences in obsession do not necessarily become visible. In the counterfactual of Galdin and Silbert (2025), when cheap signals flooded the market, employers found it harder to identify high-ability workers. For obsession to become valuable, it would need to appear in a form that others can see, and no study was found that measured this.

How much does obsession explain performance?

Traits close to obsession explain performance only moderately. Grit was defined as perseverance and passion for long-term goals (Duckworth et al. 2007). In a meta-analysis of 88 independent samples (66,807 individuals), grit was only moderately correlated with performance and retention and very strongly correlated with conscientiousness (Credé, Tynan, and Harms 2017). Of the two facets, perseverance of effort had stronger relations with outcomes than consistency of interest. The post’s person who “cannot stop fixing the hook of a single song until morning” is closer to sustained interest in one object, and that facet was the weaker predictor.

Adding passion improves prediction. In a meta-analysis of 127 studies (45,485 participants) that used the grit scale, the association between grit and performance was larger in studies whose participants were more passionate about the performance domain, and the combination of perseverance and passion predicted supervisor ratings and grades (Jachimowicz et al. 2018). The effect of practice differs by domain: deliberate practice explained 26% of the variance in performance for games, 21% for music, 18% for sports, 4% for education, and less than 1% for professions (Macnamara, Hambrick, and Oswald 2014).

Being “abnormally absorbed” has costs. Vallerand et al. (2003) divided passion into harmonious passion, an autonomous internalization of an activity into one’s identity, and obsessive passion, a controlled internalization that creates internal pressure to engage. Harmonious passion promotes healthy adaptation, whereas obsessive passion thwarts it by causing negative affect and rigid persistence.

Judgment on the obsession gap

The first half, that AI narrows ability gaps, is well supported in work with a fixed scope, but the opposite result exists for open-ended tasks such as running a business and for experienced workers on real work. No study was found that supports the second half, that differences in obsession become valuable. Traits close to obsession have moderate predictive power, and the strong form of obsession carries costs as obsessive passion. Because ability also becomes harder to see when cheap signals flood the market, it cannot be said that differences in obsession will be exposed on their own.

Passion capitalism: does what one can be absorbed in become a competitive edge?

The principle of choosing work by passion already exists

The idea that one should choose work by passion has been studied as an existing norm. Cech (2021) called the idea of centering passion in career decisions the passion principle. According to the publisher’s description, the book draws on interviews following students from college into the workforce, surveys of U.S. workers, and experimental data. It argues that the principle presumes middle-class safety nets and penalizes first-generation and working-class young adults who lack them, and that employers covet passion among job applicants but will not compensate it.

Passion can also serve as a reason to legitimize exploitation. In seven studies and a meta-analysis by Kim, Campbell, Shepherd, and Kay (2020), poor treatment of workers, such as demeaning tasks irrelevant to the job description or unpaid overtime, was judged more legitimate when the workers were presumed to be passionate about their work. Two mechanisms were at work: the assumption that passionate workers would have volunteered for the work if given the chance, and the belief that for passionate workers, work itself is its own reward.

Visible passion does not necessarily attract funding. According to the abstract of Chen, Yao, and Kotha (2009), in both a laboratory experiment and a field study, what positively affected venture capitalists’ funding decisions was the entrepreneur’s preparedness, not passion.

Is “true passion” something to be found?

The post raised the possibility that many people will discover their true passion for the first time in the AI era. Five studies by O’Keefe, Dweck, and Walton (2018) examined this idea of finding itself. People who believed that interests are fixed and must be discovered (a fixed theory) showed less interest in areas outside their existing interests, anticipated boundless motivation once a passion was found, and lost interest significantly more than people who believed interests are developed (a growth theory) when a new interest became difficult. On the other hand, according to the abstract of Chen, Ellsworth, and Schwarz (2015), both the belief that passion for work is achieved by finding the right fit and the belief that it is cultivated over time elicit different motivational patterns, but both can facilitate vocational well-being and success.

What is being valued in place of education?

No study was found showing that the value of education or experience has been replaced by passion. In U.S. payroll records, employment of workers aged 22 to 25 in AI-exposed occupations stood 19% below that of their less-exposed peers, while experienced workers showed no comparable gap (Brynjolfsson, Chandar, and Chen 2026, working paper revised August 2026, not peer reviewed). The gap arose mainly from reduced hiring of young workers. This result shows that entry-level jobs are shrinking, but it does not show that passion has come to be valued in place of education there.

Judgment on passion capitalism

The view that passion becomes a competitive edge is already a subject of sociological and social-psychological research as the principle of choosing work by passion. What the research shows is less that passion is rewarded than that passion is demanded without compensation and becomes a reason to legitimize poor treatment. The phrase “discovering one’s true passion” is close to the fixed theory that interests are found, which overlaps with a way of thinking that makes people lose interest when they meet difficulty. The research holds considerable evidence pointing in the direction opposite to the post’s optimistic reading.

The age of life prototypes: can a life be prototyped again and again?

The self as a prototype

The process of building oneself by trying things out has been described in career research. From interviews with professionals moving into more senior roles, Ibarra (1999) described adaptation to a new role as observing role models, trying out professional identities that are not yet fully elaborated, and evaluating these experiments against internal standards and external feedback, and called the selves being tried out provisional selves. The life-designing framework of Savickas et al. (2009) rests on five presuppositions about working lives: contextual possibilities, dynamic processes, non-linear progression, multiple perspectives, and personal patterns. In a design task, novices who created several prototypes before receiving feedback produced better-performing advertisements, more diverse designs, and larger gains in self-efficacy than novices who received feedback after each prototype (Dow et al. 2010). However, this result comes from a task of designing web advertisements, and whether the same happens for life choices has not been measured.

Satisfaction with changeable decisions

Being able to start over again and again can lower satisfaction. In an experiment by Gilbert and Ebert (2002), photography students who could later change which print they kept liked their chosen print less than those who could not. Even so, most participants wanted the opportunity to change. In a study that followed college students through a year of job searching, students with strong maximizing tendencies obtained starting salaries 20% higher, but were less satisfied with the jobs they obtained and experienced more negative affect during the search (Iyengar, Wells, and Schwartz 2006). However, the harm of many options as such is not consistent. According to the abstract of Scheibehenne, Greifeneder, and Todd (2010), a meta-analysis of 50 experiments (5,036 participants) found a mean effect of choice overload of virtually zero, with considerable variance between studies.

Selves not chosen do not disappear. According to the abstract of Obodaru (2012), alternative selves, meaning selves one could have been, become part of the self-concept and influence professional lives.

Judgment on the age of life prototypes

Trying out provisional selves is supported as a process of adapting to new roles. However, because changeability lowers satisfaction and unchosen selves remain, it cannot be said that being able to prototype again and again leads directly to a good life. In addition, what AI has been measured to shorten is the time to make works, not the time to try out jobs or relationships.

Judgments on the five claims

What can be said about the post’s claims?

Of the five terms in the post, dream deflation is the best supported by research: market data and experiments show that the price of making things and the value of what made things signal are falling. What is deflating, however, is the signal of being able to make things, and a premium remains for a person having spent time making something.

Instant dream death and the obsession gap are plausible as mechanisms, but their key parts have not been measured. Within the scope of this review, there was no direct research on the satisfaction of people who fulfilled a long-held dream with AI, or on differences in obsession becoming valuable. In addition, the premise that AI narrows ability gaps holds only for work with a fixed scope.

For passion capitalism and the prospect of discovering one’s true passion, the research points in the opposite direction. The norm of choosing work by passion already exists, and passion has been demanded without compensation and used as a reason to legitimize poor treatment. Life prototyping is supported as a process of trying out provisional selves, but there is evidence that changeability lowers satisfaction.

The post ends by suggesting that what AI takes from people may be their “excuses for the lives they did not live.” What research can say is that excuses work as a defense against judgments of ability (Schwinger et al. 2022); what people do once that defense is removed has not yet been measured by anyone.

This summary would need revision if the following observations accumulated. If studies that follow people who fulfilled a long-held creative goal with generative AI, comparing their expectations before fulfillment with their satisfaction afterward, repeatedly showed satisfaction smaller than anticipated, the second half of instant dream death would become supported. Conversely, if the ownership and satisfaction of people who made things with AI did not differ from those of people who made things themselves once the degree of involvement was held equal, the reading that only dreams fulfilled with AI fade would weaken.

Gaps

• Satisfaction with dreams fulfilled by AI: no study was found that compared satisfaction before and after people fulfilled a long-held creative goal (a game, a novel, a song) with generative AI. Measurements of attachment to things co-created with AI are covered in What Has Generative AI Changed About Making for Oneself? A Map of 59 Sources.
• How obsession becomes valuable: no study was found that measured whether the returns to perseverance or strength of passion rose after the spread of AI. The heritability of perseverance and self-control, and how far they can be changed, are covered in Is the Ability to Make an Effort Also a Talent? The Heritability of Grit, Self-Control, and Motivation to Learn, and How Far These Traits Can Be Changed.
• When excuses are removed: no study was found that measured how self-handicapping changes when AI removes excuses based on a lack of time or skill.
• Homogenization and the evaluation of makers: how the growing similarity of works and the premium for human-made works affect makers’ income and their willingness to continue has not been measured. Homogenization and the burden of verification are covered in AI Slop: Reading It as Outsourced Verification, Not Low Quality.
• Prototyping a life: no study was found that tested the effect of parallel prototyping (Dow et al. 2010) on career or life choices. Effectiveness studies of Designing Your Life-style interventions are not included because their bibliographic records could not be confirmed.
• Differences in starting points: whether one has the safety net needed to pursue passion (Cech 2021) connects with the discussion of initial advantages (Is the First Advantage Luck? Option Luck and Brute Luck, Ex Ante and Ex Post, and the Evidence on Initial Advantages and Fairness Judgments in the Academic Literature, Do Those Who Have Receive More? Mechanisms, Causal Identification, and Limits of Cumulative Advantage and the Matthew Effect in the Academic Literature), but it is not addressed here.

Unverified items

• No main claim remains unverified.
• Seven sources were used only within the scope of their abstracts (Van Boven and Ashworth 2007, Diener et al. 2006, Millet et al. 2023, Chen et al. 2009, Chen et al. 2015, Scheibehenne et al. 2010, Obodaru 2012). In the text, each is attributed to its authors with “according to the abstract.” For Noy and Zhang (2023), the full text of the working paper version was obtained, but the numbers are taken from the abstract of the published version.
• The full text of Cech (2021) could not be obtained, and the book is described within the scope of the publisher’s description.
• Nine sources are not peer reviewed (Peng et al. 2023, Becker et al. 2025, Kosmyna et al. 2025, Galdin and Silbert 2025, Holzner et al. 2025, Kim et al. 2026, Otis et al. 2024, Humlum and Vestergaard 2025, Brynjolfsson et al. 2026), and each is marked “not peer reviewed” in the text. Zhang et al. (2026) has a peer-reviewed version (LNCS), but its text was read in the arXiv version.
• Brynjolfsson, Li, and Raymond (2025) was read in the arXiv version (second version, November 2024), and its numbers were not checked against the published version. Brynjolfsson, Chandar, and Chen (2026) is a working paper still under revision, and its numbers are from the version of August 12, 2026.
• The correlations in Schwinger et al. (2022) are the magnitudes stated in the abstract; their direction (positive or negative) was not checked in the tables of the full text. The text reports only the magnitudes.
• The peer-review policy of the journal that published Kuehn and Corrigan (2013) was not checked on the journal’s pages.

References

All accessed on 2026-10-11.

The posts

• paji_a. (2026a, October 10). AI時代に起きる最大の悲劇は《仕事を奪われること》じゃなく… [Post on X]. https://x.com/paji_a/status/2108759364744134880
• paji_a. (2026b, October 10). さらに【ASIが実現すると人間は“夢の所有権”まで失うのかもしれないんよな】… [Post on X]. https://x.com/paji_a/status/2108816747730252020

Anticipation, attainment, and the act of making

• Loewenstein, G. (1987). Anticipation and the valuation of delayed consumption. The Economic Journal, 97(387), 666–684. https://doi.org/10.2307/2232929
• Van Boven, L., & Ashworth, L. (2007). Looking forward, looking back: Anticipation is more evocative than retrospection. Journal of Experimental Psychology: General, 136(2), 289–300. https://doi.org/10.1037/0096-3445.136.2.289
• Wilson, T. D., & Gilbert, D. T. (2005). Affective forecasting: Knowing what to want. Current Directions in Psychological Science, 14(3), 131–134. https://doi.org/10.1111/j.0963-7214.2005.00355.x
• Diener, E., Lucas, R. E., & Scollon, C. N. (2006). Beyond the hedonic treadmill: Revising the adaptation theory of well-being. American Psychologist, 61(4), 305–314. https://doi.org/10.1037/0003-066x.61.4.305
• Niemiec, C. P., Ryan, R. M., & Deci, E. L. (2009). The path taken: Consequences of attaining intrinsic and extrinsic aspirations in post-college life. Journal of Research in Personality, 43(3), 291–306. https://doi.org/10.1016/j.jrp.2008.09.001
• Kappes, H. B., & Oettingen, G. (2011). Positive fantasies about idealized futures sap energy. Journal of Experimental Social Psychology, 47(4), 719–729. https://doi.org/10.1016/j.jesp.2011.02.003
• Norton, M. I., Mochon, D., & Ariely, D. (2012). The IKEA effect: When labor leads to love. Journal of Consumer Psychology, 22(3), 453–460. https://doi.org/10.1016/j.jcps.2011.08.002
• Inzlicht, M., Shenhav, A., & Olivola, C. Y. (2018). The effort paradox: Effort is both costly and valued. Trends in Cognitive Sciences, 22(4), 337–349. https://doi.org/10.1016/j.tics.2018.01.007
• Schwinger, M., Trautner, M., Pütz, N., Fabianek, S., Lemmer, G., Lauermann, F., & Wirthwein, L. (2022). Why do students use strategies that hurt their chances of academic success? A meta-analysis of antecedents of academic self-handicapping. Journal of Educational Psychology, 114(3), 576–596. https://doi.org/10.1037/edu0000706
• Kuehn, K., & Corrigan, T. F. (2013). Hope labor: The role of employment prospects in online social production. The Political Economy of Communication, 1(1). https://polecom.org/index.php/polecom/article/view/9

Generative AI and the time to make, ownership, and motivation

• Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187–192. https://doi.org/10.1126/science.adh2586
• Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023). The impact of AI on developer productivity: Evidence from GitHub Copilot [Preprint]. arXiv. https://arxiv.org/abs/2302.06590
• Becker, J., Rush, N., Barnes, E., & Rein, D. (2025). Measuring the impact of early-2025 AI on experienced open-source developer productivity [Preprint]. arXiv. https://arxiv.org/abs/2507.09089
• Draxler, F., Werner, A., Lehmann, F., Hoppe, M., Schmidt, A., Buschek, D., & Welsch, R. (2024). The AI ghostwriter effect: When users do not perceive ownership of AI-generated text but self-declare as authors. ACM Transactions on Computer-Human Interaction, 31(2), 1–40. https://doi.org/10.1145/3637875
• Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task [Preprint]. arXiv. https://arxiv.org/abs/2506.08872
• Wu, S., Liu, Y., Ruan, M., Chen, S., & Xie, X.-Y. (2025). Human-generative AI collaboration enhances task performance but undermines human’s intrinsic motivation. Scientific Reports, 15, 15105. https://doi.org/10.1038/s41598-025-98385-2

Signaling, markets for creative goods, and homogenization

• Spence, M. (1973). Job market signaling. The Quarterly Journal of Economics, 87(3), 355–374. https://doi.org/10.2307/1882010
• Galdin, A., & Silbert, J. (2025). Making talk cheap: Generative AI and labor market signaling [Preprint]. arXiv. https://arxiv.org/abs/2511.08785
• Zhang, K., Yuan, Z., & Xiong, H. (2026). The impact of generative artificial intelligence on market equilibrium: Evidence from a natural experiment. In Web and Internet Economics (Lecture Notes in Computer Science, pp. 609–626). Springer. https://doi.org/10.1007/978-3-032-08560-3_34 (arXiv version: https://arxiv.org/abs/2311.07071)
• Zhou, E., & Lee, D. (2024). Generative artificial intelligence, human creativity, and art. PNAS Nexus, 3(3), pgae052. https://doi.org/10.1093/pnasnexus/pgae052
• Kim, S., Jin, G. Z., & Lee, E. (2026). Does generative AI crowd out human creators? Evidence from Pixiv (NBER Working Paper No. 34733). National Bureau of Economic Research. https://doi.org/10.3386/w34733
• Doshi, A. R., & Hauser, O. P. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Science Advances, 10(28), eadn5290. https://doi.org/10.1126/sciadv.adn5290
• Holzner, N., Maier, S., & Feuerriegel, S. (2025). Generative AI and creativity: A systematic literature review and meta-analysis [Preprint]. arXiv. https://arxiv.org/abs/2505.17241
• Bellaiche, L., Shahi, R., Turpin, M. H., Ragnhildstveit, A., Sprockett, S., Barr, N., Christensen, A., & Seli, P. (2023). Humans versus AI: Whether and why we prefer human-created compared to AI-created artwork. Cognitive Research: Principles and Implications, 8, 42. https://doi.org/10.1186/s41235-023-00499-6
• Millet, K., Buehler, F., Du, G., & Kokkoris, M. D. (2023). Defending humankind: Anthropocentric bias in the appreciation of AI art. Computers in Human Behavior, 143, 107707. https://doi.org/10.1016/j.chb.2023.107707
• Aguiar, L., & Waldfogel, J. (2018). Quality predictability and the welfare benefits from new products: Evidence from the digitization of recorded music. Journal of Political Economy, 126(2), 492–524. https://doi.org/10.1086/696229

Ability gaps, grit, and passion

• Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044
• Otis, N. G., Clarke, R. P., Delecourt, S., Holtz, D., & Koning, R. (2024). The uneven impact of generative AI on entrepreneurial performance [Working paper]. SSRN. https://doi.org/10.2139/ssrn.4671369
• Humlum, A., & Vestergaard, E. (2025). Still waters, rapid currents: Early labor market transformation under generative AI (NBER Working Paper No. 33777). National Bureau of Economic Research. https://doi.org/10.3386/w33777
• Duckworth, A. L., Peterson, C., Matthews, M. D., & Kelly, D. R. (2007). Grit: Perseverance and passion for long-term goals. Journal of Personality and Social Psychology, 92(6), 1087–1101. https://doi.org/10.1037/0022-3514.92.6.1087
• Credé, M., Tynan, M. C., & Harms, P. D. (2017). Much ado about grit: A meta-analytic synthesis of the grit literature. Journal of Personality and Social Psychology, 113(3), 492–511. https://doi.org/10.1037/pspp0000102
• Jachimowicz, J. M., Wihler, A., Bailey, E. R., & Galinsky, A. D. (2018). Why grit requires perseverance and passion to positively predict performance. Proceedings of the National Academy of Sciences, 115(40), 9980–9985. https://doi.org/10.1073/pnas.1803561115
• Macnamara, B. N., Hambrick, D. Z., & Oswald, F. L. (2014). Deliberate practice and performance in music, games, sports, education, and professions: A meta-analysis. Psychological Science, 25(8), 1608–1618. https://doi.org/10.1177/0956797614535810
• Vallerand, R. J., Blanchard, C., Mageau, G. A., Koestner, R., Ratelle, C., Léonard, M., Gagné, M., & Marsolais, J. (2003). Les passions de l’âme: On obsessive and harmonious passion. Journal of Personality and Social Psychology, 85(4), 756–767. https://doi.org/10.1037/0022-3514.85.4.756

The principle of choosing work by passion and its critics

• Cech, E. A. (2021). The trouble with passion: How searching for fulfillment at work fosters inequality. University of California Press. https://doi.org/10.1525/9780520972698
• Kim, J. Y., Campbell, T. H., Shepherd, S., & Kay, A. C. (2020). Understanding contemporary forms of exploitation: Attributions of passion serve to legitimize the poor treatment of workers. Journal of Personality and Social Psychology, 118(1), 121–148. https://doi.org/10.1037/pspi0000190
• Chen, X.-P., Yao, X., & Kotha, S. (2009). Entrepreneur passion and preparedness in business plan presentations: A persuasion analysis of venture capitalists’ funding decisions. Academy of Management Journal, 52(1), 199–214. https://doi.org/10.5465/amj.2009.36462018
• O’Keefe, P. A., Dweck, C. S., & Walton, G. M. (2018). Implicit theories of interest: Finding your passion or developing it? Psychological Science, 29(10), 1653–1664. https://doi.org/10.1177/0956797618780643
• Chen, P., Ellsworth, P. C., & Schwarz, N. (2015). Finding a fit or developing it: Implicit theories about achieving passion for work. Personality and Social Psychology Bulletin, 41(10), 1411–1424. https://doi.org/10.1177/0146167215596988
• Brynjolfsson, E., Chandar, B., & Chen, R. (2026). Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence [Working paper, revised August 12, 2026]. Stanford Digital Economy Lab. https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/

The self as a prototype and changeable decisions

• Ibarra, H. (1999). Provisional selves: Experimenting with image and identity in professional adaptation. Administrative Science Quarterly, 44(4), 764–791. https://doi.org/10.2307/2667055
• Savickas, M. L., Nota, L., Rossier, J., Dauwalder, J.-P., Duarte, M. E., Guichard, J., Soresi, S., Van Esbroeck, R., & van Vianen, A. E. M. (2009). Life designing: A paradigm for career construction in the 21st century. Journal of Vocational Behavior, 75(3), 239–250. https://doi.org/10.1016/j.jvb.2009.04.004
• Dow, S. P., Glassco, A., Kass, J., Schwarz, M., Schwartz, D. L., & Klemmer, S. R. (2010). Parallel prototyping leads to better design results, more divergence, and increased self-efficacy. ACM Transactions on Computer-Human Interaction, 17(4), 1–24. https://doi.org/10.1145/1879831.1879836
• Gilbert, D. T., & Ebert, J. E. J. (2002). Decisions and revisions: The affective forecasting of changeable outcomes. Journal of Personality and Social Psychology, 82(4), 503–514. https://doi.org/10.1037/0022-3514.82.4.503
• Iyengar, S. S., Wells, R. E., & Schwartz, B. (2006). Doing better but feeling worse: Looking for the “best” job undermines satisfaction. Psychological Science, 17(2), 143–150. https://doi.org/10.1111/j.1467-9280.2006.01677.x
• Scheibehenne, B., Greifeneder, R., & Todd, P. M. (2010). Can there ever be too many options? A meta-analytic review of choice overload. Journal of Consumer Research, 37(3), 409–425. https://doi.org/10.1086/651235
• Obodaru, O. (2012). The self not taken: How alternative selves develop and how they influence our professional lives. Academy of Management Review, 37(1), 34–57. https://doi.org/10.5465/amr.2009.0358