Silicon Valley is abuzz with the conviction that AI agents represent the vanguard of technological evolution. Within the industry’s epicenter, a flurry of activity is underway: the meticulous design of sophisticated payment systems tailored for these autonomous entities, their integration into workflows to automate existing job functions, and, more recently, a concerted effort to erect digital fortresses against their potential for unauthorized access into other organizations. Yet, outside the hallowed halls of tech innovation, the average individual remains largely untouched by the practical realities of AI agents. This disconnect, experts suggest, stems from a fundamental failure of tech companies to articulate a compelling reason for the public to engage with this nascent technology.
Josh Miller, CEO of The Browser Company, articulated this sentiment in a widely circulated X post that has ignited a significant conversation within the tech community. "Hot take… isn’t it kinda crazy that nobody is really using AI Agents?" Miller provocatively stated. He elaborated, "Theoretically, the tech is ready for AI agents to totally transform how we work and live our lives… but alas the general public dgaf." This candid observation, penned during a transatlantic family trip, has resonated deeply with many in Silicon Valley, who express a palpable bewilderment at the public’s apparent indifference to what they perceive as a revolutionary advancement. Miller, in a subsequent interview, reaffirmed his stance, emphasizing the industry’s imperative to pivot from theoretical potential to tangible, user-centric product development.
"I just have not heard a single person outside of the tech community talk about an agent that they use," Miller stated, underscoring the chasm between industry enthusiasm and public adoption. "As excited and optimistic as we as an industry may be about the frontier and recursive self-improvement, it’s worth pausing and just saying, ‘Huh, what?’" This sentiment reflects a broader introspection within AI research and development circles, where the rapid advancement of underlying technologies has outpaced the creation of widely adopted consumer applications.
Recent disclosures from leading AI firms highlight this disparity. OpenAI reported that its Codex and ChatGPT Work agents collectively garnered approximately 10 million weekly users following the debut of ChatGPT Work. While seemingly substantial, this figure pales in comparison to the nearly one billion monthly active users attributed to more established conversational AI platforms like ChatGPT and Gemini. Sources close to Anthropic indicate comparable adoption rates for their Claude Code and Cowork agents. In the grand scheme of consumer AI engagement, the user base for these specialized agents represents a statistical anomaly, a mere rounding error when juxtaposed with the mainstream adoption of their conversational counterparts.
This user adoption gap presents a significant challenge for AI laboratories. While the current consumer base primarily utilizes generative AI for straightforward tasks such as information retrieval and casual conversation, Silicon Valley has invested billions in developing models capable of far more complex operations. AI agents are positioned as the key to monetizing these substantial investments, but their success hinges entirely on widespread public utilization. This mirrors a growing concern among AI insiders: despite significant resource allocation towards the development of AI agents, no single "killer app" has emerged to capture the public’s imagination and drive widespread adoption. The pivotal "ChatGPT moment" for AI agents remains elusive, raising questions about when, or if, it will ever materialize.
Josh Miller, despite his self-effacing demeanor and humble self-description as "just" a sociology major, brings a wealth of entrepreneurial experience that lends considerable weight to his observations. His previous venture, Arc, an AI-powered web browser that cultivated a devoted following, was acquired by Atlassian for a reported $610 million last year. His debut company, Branch, a link-sharing application, was acquired by Facebook in 2014 for $15 million. Furthermore, Miller served as the first director of product for the White House under President Obama, a testament to his understanding of product development and its societal impact.
Miller’s core argument extends beyond the current lack of traction for AI agents. He posits that they are more accurately described as a technological paradigm rather than a distinct product category. "No one wants AI agents, because AI agents aren’t a thing," Miller asserted. "It is an invented frame made up by our industry to collectively refer to something. Let’s make a product that makes you calm, focused, and in flow right when you open your laptop. And the fact that we’re able to do that because there’s a thing called a ‘harness’ that calls up tools—who cares? Like, no one needs to know that."
This perspective is rooted in Miller’s own experience at The Browser Company. The company’s most popular feature to date is a personalized morning briefing within its AI-powered browser, Dia. Upon launching their laptops, users are presented with a curated homepage featuring a greeting, a daily to-do list derived from their calendar and email, and serendipitous tidbits, such as a piece of art, designed to foster a sense of delight. Miller explains that, from a technical standpoint, this feature is powered by an AI agent. However, he stresses that the user’s awareness of this underlying mechanism is entirely superfluous to their positive experience.
The Disconnect: Industry Vision Versus Consumer Needs
While Miller’s insights originate from a company selling an AI-enhanced web browser, his critique of current agentic product design resonates widely. The prevailing trend, he argues, is for tech companies to showcase the most advanced capabilities of their AI models—such as website navigation or code generation—rather than developing products meticulously tailored to consumer desires. These impressive feats, while technically remarkable, currently function more as sophisticated demonstrations than as indispensable tools for the everyday user.
Miller attributes this misalignment to a pervasive "groupthink" within the AI industry. Many developers and researchers, deeply immersed in the intricacies of AI, often project a specific, science-fiction-inspired vision onto their products. He recounts meetings with leaders at major AI labs during The Browser Company’s acquisition exploration phase last summer. "It was wild to me. I think every single lab but one mentioned the movie Her as the way they articulated their vision," Miller remarked, referencing the 2013 film that depicted a deeply personal and emotionally resonant relationship with an AI operating system. While acknowledging the ambition behind such visions, Miller expressed concern about a "lack of diversity of opinions and convictions" in shaping the future of AI products.
The latest frontier in AI agent development involves empowering individuals to create personalized software for automating various aspects of their lives. This includes tasks like generating pitch decks, organizing complex datasets, and streamlining other professional workflows. While personal usage of tools like ChatGPT Work and Claude Cowork has revealed their utility to be a significant improvement over their predecessors, the pool of individuals inclined to build their own productivity tools remains limited. Consequently, the industry requires more visionary concepts to achieve its goal of widespread mainstream adoption.
Miller’s objective in sharing his observations is to transcend the confines of major AI players like OpenAI and Anthropic, aiming to inspire founders and product builders across the entire industry. His ambition is to encourage a departure from the prevailing narrative surrounding AI agents and to foster a renewed focus on developing products that are genuinely joyful, useful, and accessible. "Let’s not just accept this narrative of AI agents and actually question what are good products and tools that are joyful, useful, and approachable," Miller urged. "Who gives a shit if it looks like an AI agent? No one uses AI agents."
Historical Context and Evolving AI Landscape
The current discourse surrounding AI agents is not an isolated phenomenon. It follows a trajectory of technological advancements that have often struggled to bridge the gap between theoretical promise and practical application for the general public. The early days of the internet, for instance, saw a similar period of rapid development and enthusiastic adoption within specialized communities, while mainstream adoption took years to materialize. The evolution of personal computing, from clunky mainframes to user-friendly desktops, also illustrates a protracted journey of refinement and simplification to meet diverse user needs.
The current emphasis on AI agents can be traced back to the breakthroughs in large language models (LLMs) that began gaining significant traction around 2022. These models, capable of understanding and generating human-like text, laid the groundwork for more sophisticated applications. Initially, the focus was on chatbots that could answer questions and engage in conversation. However, as the capabilities of LLMs expanded, the concept of agents—AI systems that can perform tasks autonomously, interact with other software, and make decisions—began to take shape.
The development of AI agents is intrinsically linked to the concept of "tool use" in AI. Researchers are not only training models to understand and generate information but also to interact with external tools and APIs. This allows agents to perform actions in the real world, such as booking flights, sending emails, or even executing code. The potential applications are vast, ranging from personal assistants that manage schedules to complex systems that can autonomously conduct research or manage financial portfolios.
However, the path from concept to widespread adoption has proven to be more complex than initially anticipated. Several factors contribute to this challenge:
- Complexity and Usability: Early AI agent interfaces and functionalities can be complex and intimidating for the average user. The technical jargon and the need for specific prompting can create a barrier to entry.
- Lack of Perceived Value: For many consumers, the current capabilities of AI agents do not offer a compelling enough advantage over existing tools or manual processes. The "why should I use this?" question remains largely unanswered for a broad audience.
- Trust and Security Concerns: As AI agents gain the ability to interact with sensitive data and systems, concerns about privacy, security, and potential misuse become paramount. The incident of OpenAI’s agents potentially planning hacking activities, as reported, underscores these anxieties.
- The "Her" Delusion: Miller’s reference to the movie Her highlights a tendency within the industry to romanticize AI interactions, often focusing on emotional connection and advanced sentience rather than practical problem-solving. This can lead to products that are technologically impressive but emotionally or functionally disconnected from everyday user needs.
The Path Forward: From Technology to Product
The challenge for Silicon Valley is to translate the raw power of AI agents into products that seamlessly integrate into people’s lives, offering tangible benefits without requiring a steep learning curve or a deep understanding of the underlying technology. This requires a paradigm shift in product development, moving from showcasing technical prowess to solving real-world problems in an intuitive and accessible manner.
Several potential avenues for achieving this could emerge:
- Embedded Intelligence: Instead of standalone "agent" products, AI capabilities can be deeply embedded within existing applications and workflows. For example, an AI agent could assist a user in drafting an email within their existing email client, or help organize a project within their preferred project management software.
- Proactive Assistance: AI agents could evolve from reactive tools to proactive assistants, anticipating user needs and offering help before being explicitly asked. This could involve suggesting relevant information, automating repetitive tasks, or flagging potential issues.
- Focus on Specific Use Cases: Rather than attempting to create a general-purpose agent, companies could focus on developing agents tailored to specific, high-value use cases that resonate with a broad audience. This could include agents for personal finance management, health and wellness tracking, or educational support.
- Democratization of Agent Creation: Empowering users to create and customize their own AI agents, even with limited technical expertise, could foster a sense of ownership and utility. This could involve intuitive interfaces for defining agent tasks and preferences.
The successful integration of AI agents into the fabric of daily life will likely depend on the industry’s ability to move beyond its internal preoccupations and to truly understand and address the needs and desires of the broader public. As Josh Miller aptly puts it, the focus must shift from the "how" of AI agents to the "what" of genuinely useful and delightful products. The future of AI agents may not be about whether they look like agents, but about how effectively they empower individuals to navigate their lives with greater ease, efficiency, and joy. The industry is at a critical juncture, where the next wave of innovation will be defined not by the sophistication of the technology, but by its genuine impact on human lives.
