1/9/2025
The biggest driver of change in my life has been the internet. For about half my life now, most of my routines have touched the internet in some way. In my view, it has gone from text walls on Geocities servers to our modern network of digital representations and creations, where every individual is a node.
Furthermore, the internet is not only the most massively connected network of human individuals ever assembled but also a system with self-organizing properties for information, which we, through experimentation, have put to clever uses.
In the last decade, I believe the scale of the internet has begun to be understood. Companies like Meta, Google, and Microsoft command the attention of huge networks of individuals. Several companies, like Amazon, have created entire businesses around building the middleware required for handling data processes at this scale. The implications of these attention networks, with powerful network effects sustaining them, are only beginning to be understood.
These developments include cryptographically secured token networks, digital identity concepts, privacy rights, information networks, the value of attention, and algorithms for engagement. These are all examples of the digital reality we have been slowly creating—piecemeal—by making digital signals more persistent and abstractions more tangible than ever before. We are building a fuzzy digital twin of our world, and as it increases in clarity and importance, its access and rules of engagement will become dominant public discussions.
AI is a clear reminder of the sheer importance of the internet. While certain to change the core ways we source information and interact with our environment—for example, from a query-to-query usage basis today to potentially a real-time, persistent overlay basis—AI will still likely be just one of the internet era’s miracles.
This context is helpful as we, as a people, attempt to anticipate what AI will mean for humanity. Because AI shares genetics with its internet ancestor, we are all congregating around the same types of questions. Everyone is trying to guess who will invent the “Google” of AI, the “Windows” of AI, or the “iPhone” of AI. Other questions focus on who will disrupt Microsoft or become the Netscapes, Compaqs, and other companies that failed so a few could succeed. Regarding the technology, we wonder about the stages of progress: Will “superintelligence” be attained? Do we have the hardware required to deliver more leaps in “progress,” or will progress in developing useful cases become “incremental”? Financially, we watch for signs of an AI bubble.
AI will likely evolve how individuals interact with the digital world, creating greater interconnection points between people, information, and systems. A lot of ambiguity remains about what a next-generation interface with the internet will enable. What will an exponential (and hard to model) increase in “productivity,” “efficiency,” and tangibility of the nearly infinite digital realm lead us to alter in our reality?
It is clear to me that AI will be ubiquitous, just as the internet is ubiquitous. It is not far-fetched to imagine a future where AIs outnumber people—most being digital-only, but some with great degrees of physical autonomy, mixing casually within our daily realities. This source of labor, information, and automation will become increasingly necessary as questions of individual sovereignty, privacy, and resource measurement and distribution grow more collectively important, all while our ability to manipulate reality begins to shift faster than ever before in history. I believe AI will enable a collective inflection point, allowing humanity to refocus on questions of shared significance.
One reason I believe in this future is that it is largely possible already. Many in the developed world are always connected to a network through a personal data plan. Hosting personalized AI or AIs on a dedicated private server, provided cheaply by a retail provider like Digital Ocean, could mean an always-on companion to navigate the world with. If IoT hardware manufacturers find the right use cases, proximity-based AI could explode, certainly leading to a more concrete and actionable public understanding of augmented and purely digital realities, which we might call meta-realities.
A reference point I’ve been drawn to is the digital world created in Ubisoft’s Watchdogs. Even the concept of a public network feels like something that may soon become a reality. Ideas from the “Netrunners” faction realized by CD Projekt Red’s Cyberpunk 2077 also feels to me as capturing some aspect of the seamlessness of how a hyperconnected, AI-interfaced, always-on digital world could feel. The reality enabled by human social market forces will likely be even more pervasive.
These concepts will also raise many questions about access and rules. Our digital identities were once tiny fragments of our realities. Today, the digital world is something we can walk in using a virtual reality headset. The “social laws” of the digital world will be a defining achievement once we recognize the inevitable utility of a digital network full of useful entities like AI—but also full of ways to encroach upon our personal realities and sense of sovereignty.
Unfortunately, I can’t forecast the timing, sequence, or catalyst for these events with any great certainty. Such questions, in my opinion, depend heavily on how people, as a whole, embrace and react to what becomes possible. Will we squabble for decades over semantic disagreements? Will some calamity force us to leverage our tools in new, transformational ways out of desperation?
Progress is continuous, parallel, and entangled in complex ways. It continues after we personally forget the headline from a week ago. It occurs as the result of many concurrent experiments happening all at once, whose results disseminate over various information networks. Progress also happens when one innovation spurs another in a “completely unrelated” field, transforming our world yet again.
Tactically, at the outset of the year, I am watching various hardware manufacturers relevant to these industries, such as Dell, Applied Materials, and Micron. I believe the initial phase of the “AI races” will continue to focus on finding the correct hardware stack to enable the broad type of AI ubiquity and utility described here.
The scale of compute needed to process the magnitudes of data and context we think necessary has driven waves of capital expenditure and investment in hardware R&D, all in pursuit of making leaps in output per cost of compute. Given the newness of the task at hand, while we may currently be in a “near-term” revenue contraction cycle, it seems unlikely that the final hardware stack, process stack, and optimal configuration providers have already been determined. The lack of a ubiquitous AI framework means the next generation of winners—the ASMLs and TSMCs of the “AI” age—are still to be identified. Broadly investing among existing leaders may present opportunities to profit from a narrative expansion.
There are other key hardware and technological limitations to the reality I’ve imagined. Experts point to gaps in memory efficiency, power grid constraints, the fractured nature of the internet, low relative bandwidth, huge security concerns, lack of universal signal coverage, expensive components, and countless other scaling problems. However, if we are facing scaling issues, then we have already reasonably solved for existence and determined that profits can be generated at feasibly lower costs.
I believe AI adoption will continue because its utility is too tangible to ignore. Paradoxically, AI itself may become a powerful tool for resolving many of these challenges in a mathematically unbiased way. If we achieve a world where AI overlays help us break down complex tasks with real-time instructions, the very nature of education would be questioned. An unprecedented age of discovery could unfold in fields relying on increasingly powerful computational models. With this importance will come cultural discussions about checks and controls to prevent AI—and humans—from abusing this newfound potential. However long it takes, we have set the stage for near-term iterations of human advancement.
Sincerely,
Demosthenes