Abstract
Analyses of AI power geometries have so far predominantly focused on the perspectives of Silicon Valley’s Big Tech and its “loud futuring” of AI. However, for comprehending Silicon Valley’s imaginaries surrounding AI, an analysis of the “quiet futuring” of AI-related groups and communities in the San Francisco Bay Area is equally essential because they are an integral part of its imaginative landscape of AI. Based on an ethnographic study conducted between December 2024 and February 2026, this article reveals how Silicon Valley’s imaginative landscape of AI is anchored in appropriated places, constructed within specific social figurations, and characterized by distinct discursive thickenings. Although this landscape may at first appear fragmented or even contradictory, taken together, it forms a coherent “quiet futuring” of AI. The associated visions extend beyond the technology itself, focusing instead on the potential futures it presents. Three themes dominate these visions: the possible futures of AGI, the possible futures of job loss, and the possible futures of AI geopolitics. With such an analysis, the article contributes to the discussion in “Media, Culture & Society” about the power geometries of AI by showing how Silicon Valley’s agency also refers to the local anchoring of AI’s imaginative landscape.
Introduction
In critical discussions on the “power geometries of AI” (Natale et al., 2025: 1059), imaginaries of AI are increasingly framed as products of “cyberlibertarianism,” “solutionism,” and “technofascism” (Coeckelbergh, 2026; Golumbia, 2024; Latzer, 2026; McQuillan, 2022; Morozov, 2013; Silverman, 2025). This perspective centers on the increasingly radical political views of prominent investors and CEOs such as Mark Andreessen, Peter Thiel, and Elon Musk. Their influence on the Trump administration, and their use of this influence to promote radical social ideas concerning AI, are frequently discussed in this context. While such criticism is highly important, it only partially addresses the complex “power geometries” (Massey, 1993) that shape the imaginative landscape of AI. With regard to Silicon Valley as a “global imaginary” (Marwick, 2017: 314), this landscape is far more deeply anchored locally than personalized attributions suggest. It represents a fabric of diverse groups that extend well beyond the transhumanist and longtermist circles referred to by Gebru and Torres (2024) as the “TESCREAL bundle,” such as Effective Altruism and Rationalism.
With such groups in mind, I would like to substantiate the following thesis: while critical analyses of Silicon Valley and AI have largely focused on the “loud futuring” of radicalizing investors and CEOs, the “quiet futuring” of AI-related groupings in the San Francisco Bay Area is at least equally crucial for understanding the imaginaries surrounding AI. In both cases, the emphasis lies on the promise of possible digital futures with AI (Hepp, 2025, 2026a). However, whereas “loud futuring” is primarily directed at investors, politicians, and the general public, “quiet futuring” is more oriented toward individual communities and groups, establishing a long-term horizon for imagining AI. Both are clearly interrelated: the visions of AI that emerge in “quiet futuring” serve as a reservoir of ideas that are often condensed and amplified through “loud futuring,” thereby achieving wider circulation. At the same time, “loud futuring”—for example, CEOs’ public statements about timelines of Artificial General Intelligence (AGI)—contributes to structuring the agenda of “quiet futuring” by stimulating certain topics. In this bidirectional relationship, actors engaged in “quiet futuring” tend to present themselves as apolitical, even though their visions of AI are highly political in nature. In this way, they contribute to the power geometries of AI.
This argument can be illustrated with an example from my ethnographic field research. In mid-September 2025, Jasmine Sun, a San Francisco member of the pioneer community Reboot, posted on Substack asking her readers which article she should publish: a “serious political” one or a “silly tech thing.” In the Substack comments, her readers initially expressed a clear preference for the “silly tech thing,” sparking a debate about whether there was anything meaningful left to say about politics. Sun reflected further on these discussions and voiced her concern that her tech circles in San Francisco were not taking politics seriously enough, noting that people there seemed rather insulated from it. While some participants suggested that this detachment might, in fact, be the real issue, Sun ultimately decided to publish the “silly tech thing.” In her article, she then describes how the “AI gold rush has sparked a vibe in San Francisco,” but also points to the “cracks” within this atmosphere: although belief in technological progress as a driver of history, the cult of the brilliant individual, and a deep faith in markets are widespread, these elements do not fit into a coherent ideology.
In this example, we are dealing with a form of “quiet futuring” that does not neatly fit into the overarching triad of “cyberlibertarianism,” “solutionism,” or “technofascism.” In some respects, it even constitutes a critique of these very narratives. Its intended audience is not the general public, but rather “technologists,” particularly those based in the Bay Area. From a social-scientific observer’s perspective, however, it is precisely in the call for the “silly tech thing” that the underlying politics of AI in the San Francisco Bay Area become apparent—namely, the tendency to imagine AI not only as a future market force for societal change, but also to frame this very process as apolitical.
To grasp this, it is necessary to situate “quiet futuring” within the wider imaginative landscape of AI. By imaginative landscape of AI, I refer to the locally grounded ensemble of visions, positions, and conflicts that have emerged in relation to AI—particularly the relatively novel ones that have developed alongside this emerging technology. This imaginative landscape is not confined to technological aspects but encompasses broader ideas and concepts regarding the future of communication and society with AI, as well as possible risks to humanity. The current imaginative landscape of AI is far from uniform; it is fragmented, diverse, ideologically charged, and highly contradictory.
What characterizes the imaginative landscape of AI in the San Francisco Bay Area? And what can we learn from its analysis for broader questions of AI-related futuring? I address these research questions through a multisited media ethnography conducted in the Bay Area between 2025 and 2026. I begin by examining the concept of the imaginative landscape and the current state of research on this topic. I then outline my ethnographic approach. This leads to a threefold argument: an imaginative landscape is anchored in appropriated places, constructed within specific social figurations—such as the aforementioned pioneer community Reboot or longtermist and transhumanist groups—and characterized by distinct discursive thickenings. As I argue in the conclusion, these elements together constitute a form of “quiet futuring” that currently permeates the discourse in Silicon Valley and is a crucial part of AI’s power geometries. With this analysis, the article contributes to the discussion in “Media, Culture & Society” about the power geometries of AI by showing how Silicon Valley’s agency also depends on the local anchoring of AI’s imaginative landscape. 1
Approaching the imaginative landscape of AI
Within media and communication research, there has been a longstanding discourse on the imaginations surrounding AI. This debate, building on earlier work (e.g. Jasanoff and Kim, 2013; Taylor, 2004), focuses on two types of AI-related imaginations: first, the visions of AI held by individual corporate actors, such as tech companies, and by collective actors, such as communities; and second, the imaginaries of AI that emerge within and across societies.
At the level of corporate and collective actors, previous research has examined how specific visions of AI relate to the characteristics and interests of different types of actors. For example, lawmakers’ visions of AI are shaped by their understandings of regulation (Bakiner, 2023), corporations like Meta/Facebook develop visions of a “better world” linked to AI (Haupt, 2021), and pioneer communities in journalism envision a future newsroom shaped by AI (Hepp, 2025). Another area of interest is transhumanist and longtermist groups, such as Effective Altruism and Rationalism, which associate AI with the prospect of an emerging superintelligence that could pose an “existential risk” to humanity (Gebru and Torres, 2024; Kessler, 2024).
At the societal level, there is a wide-ranging debate on how comprehensive sociotechnical imaginaries of AI relate to specific technology policies and, therefore, the power geometries of AI (Natale et al., 2025). While many studies focus on the emergence and dissemination of AI imaginaries within societal discourses (e.g. Bareis and Katzenbach, 2022; Bory et al., 2025; Scott Hansen, 2022), others address more targeted research questions. For example, they examine the extent to which AI imaginaries reinterpret existing understandings of (techno)capitalism (e.g. Berlinski et al., 2024), the kinds of AI-based healthcare that are envisioned (Hoff, 2023), or the design of national AI governance structures (e.g. Mao, 2025).
Such research demonstrates that, on the one hand, the exploration of AI imaginaries is a well-established field of study. On the other hand, the terminology used so far suggests that only a few studies have addressed how different visions relate to one another, particularly without assuming a unified imaginary that exists for—or even beyond—society. Empirical studies emphasize that related imaginaries are “multiple, contested, and commodified” (Mager and Katzenbach, 2021: 226). It has also been rightly noted that individual “visions” of corporate and collective actors do not necessarily translate into societal “imaginaries” (Hepp, 2026a; Hilgartner, 2015: 33).
However, the conceptual and methodological implications of such complexities have received far less attention. This is particularly the case when such imaginings are understood as an integral part of the power geometries of AI (Massey, 1993; Natale et al., 2025: 1059). This article seeks to address this gap by analyzing the San Francisco Bay Area’s imaginative landscape of AI—the locally anchored ensemble of visions, positions, and conflicts that emerge in relation to AI and that is closely related to its power geometries.
To approach this landscape, it is helpful first to consider how earlier social-scientific concepts of the “imaginary” have been understood. Arjun Appadurai, for example, spoke of “imaginary landscapes” that shift in the context of globalization. He described these landscapes as shaped by the reproduction of images, the creation of imagined communities, and the construction of collective aspirations (Appadurai, 1996: 31). Building on this, the concept of the imaginative landscape allows us to develop a more integrative view of the imaginations connected to particular digital technologies and digital futures. To grasp a “landscape” (Antrop, 2018; Bolin, 2006; Kühne, 2019), one requires—as was emphasized early in cultural and social geography (Granö, 1929)—a Fernsicht (“distant view”) rather than a Nahsicht (“close view”). Taking this broad perspective on the imaginative landscape of AI, its infrastructures, and the associated visions of digital futures, it becomes clear that at least three interconnected levels should be considered: appropriated places, social figurations, and discursive thickenings (Hepp, 2026b).
Imaginative landscapes are anchored in specific appropriated places. This claim draws on research into the material dimension of landscapes: although imaginative landscapes encompass a holistic perspective on individual projections, collective visions, and societal imaginaries, they remain anchored in specific locations—places where people gather, where laboratories are situated, and where they live. If we take the sociomateriality of imaginings seriously (Suchman, 2012), research on imaginative landscapes cannot ignore their grounding in appropriated places.
Imaginative landscapes emerge within social figurations. They are actor-bound precisely because they involve power and ideology (Balbi, 2023). This applies particularly to corporate actors, such as companies or public authorities, as well as collective actors, including communities or movements. From the perspective of process sociology, such supra-individual actors are themselves “figurations” (Elias, 1978: 13), that is, they are characterized by specific actor constellations, shared frames of relevance, and established practices. These figurations are continuously being made and remade. Consequently, an imaginative landscape is articulated through a “figuration of figurations” (Couldry and Hepp, 2016: 57), representing the overarching figuration formed by different actors in their shared orientation toward one another, for example, when negotiating what defines AI.
Imaginative landscapes generate discursive thickenings. The concept of imaginative landscapes emphasizes that they are not homogeneous at the discursive level. They encompass a range of positions, arguments, and visions of the future, some of which may be directly opposed to one another. In this sense, imaginative landscapes of AI refer to the entirety of visions, positions, and conflicts related to AI. They extend beyond technological aspects to include broader ideas about the future of communication and society with AI, as well as potential risks for humanity. While imaginative landscapes do not resolve into a single, homogeneous perspective and contradictions persist within them, they produce something else: various “thickenings” (Löfgren, 2001: 11) of discourse, which, through their contradictions, delineate a kind of “horizon” (Schütz, 1962: 80) that shapes what can be articulated and, consequently, which futures can be imagined at a given point in time.
Such a distinction between the three layers of imaginative landscapes should not lead us to treat them as separate from one another. On the contrary, the layers reflect different aspects of a single landscape, which can only be fully understood by examining their interplay. Likewise, the potentially more spatial distinction between layers should not lead us to overlook the temporal dimension. Imaginative landscapes are processual phenomena and should always be considered in terms of their emergence and ongoing development.
Multi-sited ethnography and extended situational analysis
To capture an imaginative landscape of AI as outlined above, a complex, multi-level approach is required. The methodology underlying my arguments combines “multisited ethnography” (Marcus, 1995) with a further development of “situational analysis” (Clarke et al., 2022) within the framework of Grounded Theory. Multisited ethnography is particularly suitable for this research because reconstructing an imaginative landscape entails conducting interviews and observations with numerous actors operating across diverse sites. Integrating situational analysis further allows for the inclusion of the physical locations of these actors as well as the discourses that circulate within the landscape.
Specifically, my analysis is based on two stages of data collection. In December 2024, an exploratory phase was conducted through a well-connected contact in the San Francisco Bay Area. The sampling strategy combined purposive and snowball sampling: initial participants were selected to represent a range of actor types—bloggers, journalists, and scientists—with subsequent interviewees identified through referrals and field observations. These interviews were analyzed in a first step, focusing on field exploration and identifying the various groupings that would become the focus of subsequent empirical research.
To this end, a second data collection process took place between June 2025 and February 2026. This ethnographic phase included three field trips to the Bay Area, supplemented by online-based research. Here, the approach shifted to “theoretical sampling” (Glaser and Strauss, 1999: 45) in the sense of Grounded Theory; that is, cases were added during data analysis in order to gradually expand and deepen the categorization. Particular attention was given to locations, events, and communities identified as influential. All interviews and observations were documented in memos. This fieldwork was complemented by a “netnography” (Kozinets, 2020) of online forums related to these communities, as well as in-depth online interviews. All interviews were transcribed automatically and subsequently reviewed by student assistants.
Table 1 below provides an overview of the data collected. Researchers were consulted only during the exploratory phase to gather field information—their statements are not considered as primary sources in the subsequent analysis. Community members are defined by the fact that they feel they belong to a specific community. Organizational representatives represent a specific company, lab, or university. Technologists are independent developers, founders, and engineers who did not primarily identify with a specific community but contributed to the landscape through their professional practice. In the observations, the locations refer to various buildings and spaces, while the events refer to conferences, meetings, and protests.
Data from multi-sited media ethnography.
For data analysis, a combination of open coding and an extended situational analysis was employed, wherein this article focuses on the latter. I refer to my approach as extended situational analysis because it goes beyond the original ideas of Clarke et al. (2022). In contrast to Marres (2020) my aim is not to further develop this into “situational analytics” in the sense of digital methods, but rather to move beyond the original focus on “situations” within “social worlds” and “arenas” to arrive at a mapping that is more closely tied to an overall landscape.
In other words, the expanded situational analysis presented here aims to capture the entire imaginative landscape of AI in the San Francisco Bay Area, taking into account not only the physical locations but also various figurations of actors and discourses, and to understand these in terms of their relationality. This mapping was developed iteratively, drawing on memos from interviews and field observations while simultaneously refining the theoretical concept outlined in the previous section as its basic structuring framework.
Appropriated places
A notable feature of recent AI advancements in the Bay Area is their strong focus on San Francisco. While tech giants such as Alphabet/Google, Apple, and Meta/Facebook are located between San Mateo and San José—the traditional geographic core of Silicon Valley—current AI developments are predominantly concentrated in San Francisco. According to the interviewees, this trend can be attributed to two mutually reinforcing factors. First, San Francisco hosts prominent AI “labs” such as OpenAI and Anthropic, which play a central role in advancing generative AI and shaping public discourse around it. Second, the current generation of “technologists” 2 tends to prefer living in vibrant city neighborhoods rather than the more suburban areas of lower Silicon Valley. While San Francisco has long been the cultural hub of technology-related communities in the Bay Area (Marwick, 2013; Turner, 2006), it now also serves, at least economically, as the center of emerging AI companies.
The significance of such urban spaces is illustrated by the example of OpenAI’s new headquarters. While its interior is intended—at least in terms of self-presentation—to symbolize a place for reflective engagement with AI (represented among others by an elaborately designed library), the two OpenAI buildings in San Francisco remain contested sites. StopAI, a group of radical AI opponents, has regularly staged protests at these locations. Remarkably, the core members of StopAI are not originally from the Bay Area but were sent there by a donor with $200,000 to organize protests against A(G)I. And even OpenAI’s former headquarters in the historic Pioneer Building continues to be symbolically contested: In October 2024, Musk’s xAI Corp. moved into these premises, effectively staking a symbolic claim to the position OpenAI had occupied in AI development.
However, it is not only the lab locations that anchor the imaginative landscape of AI. A diverse array of institutions also contributes to current AI development, with Stanford and Berkeley playing particularly crucial roles. Their AI programs, research institutions, and initiatives such as Stanford’s Human-centered Artificial Intelligence (HAI) and Berkeley Artificial Intelligence Lab (BAIR) serve as important academic hubs for ongoing developments. Nevertheless, within the field itself, this centrality is also somewhat downplayed. Members of various pioneer communities and developers I interviewed perceive these two universities as more peripheral, believing that the real activity takes place in the “labs.” In their view, only private corporations can “scale” current AI technologies to a level that enables the necessary development push. From the perspective of university-based interviewees, this reflects a division of labor that has already proven effective in cases such as Google.
Other important elements of local anchoring are the various houses and collaborative spaces. Unlike purely commercial co-working locations, these spaces combine technology development with community building. They offer a unique mix of living quarters and venues for hackathons, developer meetups, and investor events. Many residents focus on founding AI-related startups, earning these spaces the nickname “start-up houses.” This dynamic recalls the Web 2.0 era in San Francisco and is reflected in public reporting (see, e.g. Rocha, 2025; Stone, 2025.). Notably, the AGI House stands out. The first AGI House originated from the Neogenesis House in Hillsborough, near San Mateo, and one of its founders and initial tenants later established another AGI House in San Francisco. A legal dispute over branding ultimately underscores the symbolic significance these places hold within the imaginative landscape of the Bay Area. The underlying accusation is that each party is attempting to capitalize on the concept and its associated name.
Spaces differ from houses in that they are not designed for communal living; however, community building remains important. A prominent example in the Bay Area AI landscape is Frontier Tower, established in 2025 with roots in the crypto community. It features an AI-dedicated floor, alongside floors for “Accelerate!,” “Longevity,” and “Ethereum,” serving as the hub for the Bay Area Ethereum community. Other notable AI spaces include MOX, closely associated with Effective Altruism, and Constellation, which focuses on “AI safety” and offers fellowships. Each of these spaces reflects a distinct orientation and community within the AI landscape, providing work and learning areas, desks, conference and break rooms for independent developers, researchers, and startups, and, according to their mission statement, spaces for authors and artists. Ultimately, these spaces function as central hubs for what could be termed “independent AI technologists.”
Other places also play a significant role, even if they are not immediately associated with the AI landscape. These include venues where important events and gatherings regularly take place. This is not limited to large conference centers in San Francisco but also encompasses smaller venues such as The Commons in Hayes Valley, a district increasingly becoming a central hub for AI innovators. The Commons positions itself on its webpage as a “fourth place” (beyond home, work, and public spaces) and hosts both its regular meetings and events organized by groups such as the pioneer community Reboot. According to the interviewees, after the coronavirus pandemic, such “fourth places” became particularly important for younger “technologists.”
In Berkeley, another notable place of the AI landscape is the “campus” Lighthaven, operated by Lightcone Infrastructure. Lightcone Infrastructure is a key institution for Effective Altruism and Rationalism in the Bay Area and manages the LessWrong forum. Lighthaven hosts smaller regular events such as a LessWrong reading group and larger conferences, including The Curve in 2024 and 2025. This conference describes itself as a “conference where thinkers, builders, and leaders grapple with AI’s biggest questions” (https://thecurve.goldengateinstitute.org (accessed: May 7th, 2026)).
Another major event is the annual EAGlobal, the flagship conference of Effective Altruism in San Francisco, organized by the Centre for Effective Altruism (CEA). Officially, Anthropic and, even more strongly, OpenAI have distanced themselves from Effective Altruism, but the conference in February 2026 paints a picture of close interconnections: not only are MOX and Lighthaven helping to organize the conference’s pre-events, with MOX hosting an after-party, but also a large number of attendees—45 people out of a total of around 1200 attendees—are closely associated with the MATS Program (Machine Learning Alignment Theory Scholars), a Berkeley-based fellowship program that brings “AI safety” researchers together with experienced mentors.
Important for understanding the relationship between Effective Altruism and the AI Labs, however, is the fact that a remarkable number of participants are from Anthropic (current employees: 26; former/indirectly affiliated: 12) and OpenAI (current employees: 6; former/indirectly affiliated: 10). The topics these two groups represent are, in particular, “AI safety.” Among the participants from Anthropic were not only their recruiting team (which had its own booth), but also, in addition to technical staff, employees responsible for Claude’s values and character training—the work underpinning what Anthropic officially calls the “Model Spec” and is colloquially referred to as Claude’s “Constitution.” At the conference, the dominant topic was not whether, but when, a superintelligence would emerge. In the interviews conducted with participants, it became clear that this very belief is a driving force behind their own commitment to “AI safety,” and also behind financial investments.
The places (spaces, houses, etc.) and events (parties, conferences, etc.) presented so far represent only a selection. Nevertheless, they illustrate the extent to which Silicon Valley’s imaginative landscape of AI is locally anchored. Its debate unfolds in specific locations, with each site reflecting a distinct orientation.
Social figurations
The actors mentioned so far already point to the second layer of the imaginative landscape, namely the diverse social figurations. This layer encompasses more established “corporate actors,” such as companies, universities, and similar entities, as well as looser “collective actors,” including pioneer communities, tech movements, and groupings that may be legally recognized as non-profits. Together, these supra-individual actors form a larger figuration within which the imaginative landscape of AI in the Bay Area is actively negotiated.
This overarching figuration is a theme that emerged repeatedly in my ethnographic research. Actors in the field themselves recognize the coexistence of different “communities” or “camps.” Both terms were used by the interviewees, depending on whether they emphasized the social aspect or the different ideological orientations. They often emphasize that, despite this coexistence, commonalities between these groups—such as shared concerns about “AI safety”—are frequently underappreciated, and that on specific issues, the groups may be closer to one another than their public positioning suggests. To address this, individual non-profits, such as the recently established Golden Gate Institute for AI, have set out to build bridges between these “camps.”
To comprehend this layer of social figurations, it is crucial to recognize the interdependence of different actor types. A key distinction exists between “corporate actors,” whose shared objectives and practices are formed through “binding agreements” (as in corporations and research institutions), and “collective actors,” whose shared objectives and practices develop through “mutual observation” (as in pioneer communities and movements; Schimank, 2010: 329). This analytical framework shows that, within the Bay Area’s imaginative landscape of AI, the relevant collective actors intersect horizontally with the actor constellations of corporate actors.
The significance of this becomes clear through specific examples. Key collective actors within the longtermist and transhumanist groupings include Effective Altruism and Rationalists, while pioneer communities such as Ethereum and Reboot also play a prominent role. Among corporate actors, OpenAI and Anthropic—referred to by interviewees as “labs”—stand out. Universities such as Berkeley and Stanford, with their associated research institutions, also play a significant role. With the growing “hype” surrounding AI (Bareis et al., 2023: 10; Hepp et al., 2023a: 41), other actors such as the Golden Gate Institute, the Collective Intelligence Project, and critical groups like Stop AI have increasingly aligned with this core constellation, positioning themselves in relation to the labs.
Examining the emergence of this overall figuration from a historical perspective highlights the significant influence of longtermist and transhumanist groupings on developments at OpenAI and Anthropic, as is also documented by the already discussed EA Global conference. In addition to such events, many members of these “labs” were involved in discussions on and around LessWrong, and the argument that the possible emergence of a superintelligence constitutes the greatest “existential risk” for humanity shaped both internal deliberations and the public presentation of OpenAI. At this point, the statements of the interviewees align with Hao’s (2025) descriptions of AI “boomers” and “doomers.” Later, concerns that safety research was being deprioritized relative to capability scaling contributed to the founding of Anthropic by former OpenAI employees. At the same time, Effective Altruism extended its influence through universities, for instance, by specifically “recruiting” high-achieving young people via scholarships. The blog 80,000 Hours also recommended pursuing a degree in AI as a way to make the greatest possible contribution to mitigating humanity’s “existential risks.” Some interviewees reported that this blog, along with other Effective Altruism publications, not only drew them to AI but also motivated their move to the Bay Area.
In a sense, Effective Altruism—originally conceived as a utilitarian-oriented movement and extensively funded between 2020 and 2022 by the crypto industry (FTX, Sam Bankman-Fried)—can be considered a pioneer community in AI development. Many visions of potential AI futures and associated risks were imagined within it long before they became technically feasible. Pioneer communities have also emerged directly from the field, as exemplified by the previously mentioned Reboot community. What distinguishes this community is its founding during the coronavirus pandemic by young “technologists” aiming to create a new home for “techno-optimism.” Both terms were repeatedly emphasized by interviewees during conversations with community members. A particular focus, though not the only one, is on AI and its transformative impact on society (Hepp, 2026a). Discourse takes place on Substack, a Discord channel, and in the magazine Kernel. The role of this pioneer community as an “intermediary” (Bourdieu, 2010: 151) is evident: it comprises not only young individuals working in the tech industry, including those employed by “labs” such as Anthropic or conducting AI research at universities, but also rising tech journalists and those actively involved in tech activism. Consequently, the community’s discourse exhibits a highly mediating character, with personnel overlaps with initiatives such as the “Collective Intelligence Project.”
Ethereum, a crypto pioneer community founded in 2014, is also gaining significant traction within the AI landscape. Initially a key player in the crypto space, its visionary founder, Vitalik Buterin, recognized the potential link between crypto and AI risk as early as 2016 and actively sought to bridge these discourses (Buterin, 2022b). Beyond its involvement in projects such as the Frontier Tower, Ethereum has made AI and its potential applications in crypto a central focus of recent activities. The community’s role as an intermediary is particularly noteworthy. Through its association with concepts such as the network state, it not only promotes projects such as pop-up cities and villages where young AI enthusiasts gather but also engages with cyberlibertarian and right-wing discourses in the Bay Area. 3
There is an ideological affinity between Ethereum’s blockchain architecture and anti-state ideologies, in the idea that code can replace law, that smart contracts render judicial bodies obsolete, and that decentralization is inherently democratic and systematically undermines state institutions (Brody et al., 2023; Golumbia, 2024; Hepp, 2026b). Paradoxically, this stands in contrast to Ethereum’s emphasis on open structures. At this point, there is a close connection with the Mozilla Foundation’s concept of “public AI,” which is also discussed by Reboot. This connection is further evidenced by the Ethereum Foundation’s support for MozFest 2025, the Mozilla Foundation’s main annual event.
Discursive thickenings
The term discursive thickening is deliberately chosen to highlight that AI discourses are not characterized by homogeneous visions or even a uniform imaginary. Rather, they form around shared themes, enabling diverse positions and potentially contentious debates. Discursive thickenings are inherently temporary, existing for a limited period before dissolving, merging with other thickenings, or disappearing entirely.
During my field research in the San Francisco Bay Area, three distinct discursive thickenings emerged as prominent threads connecting various encounters, observations, and readings: (i) the arrival of AGI, (ii) job losses, and (iii) the geopolitics of AI.
The significance and controversy surrounding the arrival of AGI are best illustrated by a field observation. 4 In June 2025, I attended the inaugural event of the newly established Golden Gate Institute for AI. During small talk before the main discussion, attendees gathered at tables for drinks and finger food. At my table, the conversation focused on when AGI might arrive. Opinions ranged widely: one extreme predicted its arrival within 3 years, while only a visiting engineer from India firmly believed that AGI would never materialize.
This intense focus on the near-term future of AGI as “superhuman” and “transformative” was further fueled by the launch of the website ai-2027.com by the AI Futures Project and Lightcone Infrastructure, which repeatedly came up in discussions. Similarly, during Stop AI protests in front of OpenAI’s main building, flyers addressed the imminent arrival of AGI and the potential dangers it poses to humanity.
In contrast, an interview with a pragmatist at one of the AGI houses focused on the commercial prospects of AI, largely setting aside the risk discourse. In these conversations, AGI was often perceived as a multi-purpose, cross-domain AI whose imminent arrival represented a lucrative business opportunity.
Therefore, despite the lack of consensus in the Bay Area regarding the timeline or nature of AGI, the theme remained central. In other words, while there were numerous distinct visions of AGI, far from a uniform imaginary, there was remarkable consistency in viewing its potential arrival as a key issue.
The second discursive thickening centers on job losses. This discourse extends well beyond research labs and technologists, becoming a prominent theme in public discussions and advertising around AI. For example, in summer 2025, drivers in San Francisco frequently encountered advertising posters promoting AI applications, featuring slogans such as “stop hiring humans for work they hate” to market new “AI Employees” (artisian.co), alongside messages like “don’t replace” (outset.ai) humans, advocating for AI-mediated research platforms. Recurring events also focus on this topic, such as the CalMatters’ “AI and the Future of Work in California,” which addressed the impact of AI on low-wage and tiered workforces and how neural AI technology is reshaping the workplace. This event took place in July 2025 at the offices of The James Irvine Foundation, which also sponsored it.
The discourse widely entertains the notion that AI could “replace” white-collar jobs, including software development, as reflected in interviews with developers and founders. Imaginings also extend to policy responses, such as a “universal basic income” (UBI): if AGI were to take over a significant portion of jobs, humans would need alternative means of earning a living, both for individual sustenance and to maintain capitalist structures. The seriousness of these considerations is further evidenced in Sam Altman’s investment in research on this topic 5 and in representatives actively engaging in discussions about such potential futures. 6
A third discursive thickening of the evolving imaginative landscape of AI in the San Francisco Bay Area is the geopolitics of AI. In the social sciences, geopolitics typically refers to the spatial foreign policy actions of major powers within the framework of geostrategy (Kovac, 2023). However, when it comes to AI geopolitics, a striking contradiction stands out in the San Francisco Bay Area: On the one hand, “technologists”—as is evident in the example of Jasmine Sun and Reboot cited at the beginning—like to present themselves as apolitical. On the other hand, the geopolitical issue of the so-called “AI race” between the U.S. and China is a constant topic of discussion.
Jasmine Sun, to continue with her example, processed her own stay in China—her parents’ country of origin—in a reflection on the rapid technological development and the “abundance” of living there (see https://jasmi.news/p/china-2025 [accessed: May, 7th, 2026]). But this “race” was also a recurring underlying theme in other conversations. In part, this concerns issues related to the U.S. export control regime for semiconductors (CHIPS Act, BIS regulations). At its core, however, the issue revolves around broader questions, namely the capacity for innovation and the role that technologies are said to play in shaping society. And here, “AI safety” remains a key issue: Who is more likely to be trusted to develop a “friendly” or “aligned” AGI—Anthropic in San Francisco, where this is a topic of conversation in the hallways, or DeepSeek in Hangzhou, where belief in the possibility of AGI is far less widespread?
These examples show that, when considering the imaginative landscape of AI, it is crucial to broaden our understanding of geopolitics beyond the boundaries of nation-states. Companies, particularly those involved in AI, are also competing for global supremacy, hoping to become “empires of AI” (Hao, 2025) in their own right. On one hand, there is the idea of resisting these empires, as exemplified by the previously discussed StopAI, whose advocacy is fundamentally oppositional. For example, in August 2025, StopAI organized protests together with the Land is Life movement in front of Scale AI’s headquarters, viewing the company as an accomplice of Palantir and OpenAI. On the other hand, interviews with founders and technologists explore questions of global governance in a more constructive light, addressing how AI technology should fit into the existing digital capitalism of Silicon Valley. Other “technologists” often argue that AI is such a key technology that it requires a more “public” or “open” approach. The level of discussion on these issues ranges from achieving public transparency of language models and training data to regulating AI for the public good. These positions have also gained traction in Californian politics. For instance, Governor Gavin Newsom stated in an interview that he aims to regulate AI as a key technology and establish a public fund for it (https://www.youtube.com/watch?v=p_ftESFGNW4 (accessed: May 7th, 2026)).
Conclusion: The “quiet” and “loud futuring” of AI
If we now synthesize the previous observations regarding the imaginative landscape of AI in the San Francisco Bay Area, several fundamental mechanisms of “quiet futuring” and its circular relationship to Big Tech’s “loud futuring” become clear.
First and foremost, this concerns the ambivalence of the imaginative landscape itself. Also in the San Francisco Bay Area, we can observe a certain contradiction, in the sense that AI visions do not conform to a single ideology. If ideology is defined as “a particular definition of reality [that] comes to be attached to a concrete power interest” (Berger and Luckmann, 1966: 141), it becomes crucial to identify the multiple ideologies and power interests at play. As my analyses have shown, ideas of cyberlibertarianism are indeed present in the imaginative landscape of AI, as well as right-wing concepts such as those associated with the network state.
But these elements do not cohere into a unified ideological structure in which all efforts aim to expand the power of one form of authoritarianism marked by the convergence of state and corporate empires in relation to AI. Other perspectives remain, such as visions of “public AI.” This becomes particularly evident when the “future of collaborative technology and democracy” is envisioned in a new “plurality” of AI (Weyl and Tang, 2023), offering possibilities beyond the framework of the so-called “Californian ideology” (Barbrook and Cameron, 1996; Hepp et al., 2023b). This is evident not only in the various appropriated places but also in the diverse social figurations, ranging from pioneer communities such as Ethereum and Reboot to labs such as OpenAI and Anthropic.
At the same time, however, the discursive thickenings of this landscape also show to what extent this imaginative landscape is about futuring: the arrival of AGI ultimately raises the question of at what point society’s future will be fundamentally shaped by an artificial intelligence that transcends human capabilities. The issue of job losses concerns the future economy of a society fundamentally defined by AI, and the geopolitics of AI address the question of who can shape this future and how. In “quiet futuring,” such discourses are first of all addressed to specific group members, supporters, and interested parties. Communication with them also takes place at closed events, via community media such as specialized magazines, Substacks, and online forums, or through specific social media networks.
From this perspective, however, Effective Altruism and the Rationalist Community appear to be particularly influential driving forces. Although they are frequently criticized by other actors in the landscape (e.g. Triedman and Ahlawat, 2022), many fundamental temporal arguments originate from their circles: only if one adopts a long-termist view of humanity and assumes that AI could give rise to a superintelligence does it become reasonable to regard AI as the greatest possible “existential risk” to humanity. And it is precisely this perspective that not only drives a futuring discourse—one that takes concrete form in various visions of when superintelligence might emerge such as AI2027—but can also motivate corporations to adopt a particular approach to “AI safety” or, conversely, to fundamentally oppose AI, as is the case with StopAI (see, e.g. Gleiberman, 2023; Young, 2024).
Such nexuses are what make the imaginative landscape of AI in the San Francisco Bay Area both distinctive and analytically challenging. Many of Big Tech’s “loud futuring” statements are difficult to properly contextualize without taking into account this “quiet futuring,” which serves as their ongoing reservoir of ideas. This also applies to critiques of techno-fascism: there is no question that a shift to the right is currently taking place in Silicon Valley (Golumbia, 2024; McQuillan, 2022; Silverman, 2025). In particular, Big Tech’s “loud futuring” is increasingly characterized by positions that, especially with regard to AI, aim for a “fusion of state, corporate, and paramilitary powers into a loosely organized yet effective apparatus of domination” (Coeckelbergh, 2026: 4). However, rather than simply criticizing such a shift in general terms based on isolated statements by CEOs, it is essential to view these statements as part of locally anchored imaginative landscapes. Power geometries can hardly be analyzed without this.
Footnotes
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This article is based on research conducted in the DFG/FWF funded Research Unit 5656 “Communicative AI,” project “Pioneer Communities: Imagining ComAI and its possible futures” (funded by DFG, German Research Foundation—516511468).
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.