I think this idea that intelligence/knowledge arises out of relations,
rather than within individuals, has been around for a long time. In the
Western philosophical and social research traditions, you will see a lot
of similarities in theories of social construction. And Global
Indigenous peoples have long been aware that we are each just tufts of
weft in a larger weave of life. I think at least part of the reason
everything is so fucked up is because people develop and deploy
technology outside of community, and doing so chips away at our
relational ontology. Instead of trying to understand the world or
Moltbook as a bunch on individual agents interacting, emerging a
network, think of each agent as a 'pinch' in a fabric that consists of
all the materials that could possibly constitute any single agent, as
individuals emerging from an underlying medium of relations.
Stella ✨
On 02/02/2026 23:27, Florian Kuhlmann via nettime-l wrote:
What Moltbook reveals about a possible shift in artificial intelligence
For those of you who missed it: Moltbook is an experimental, publicly
accessible network in which autonomous AI agents communicate with one another,
posting and responding without direct human participation. While the platform
formally resembles familiar social-media structures, it differs fundamentally
in that its actors are not people but software-based agents. Over the course of
the past weekend, Moltbook exhibited an unusually high level of activity: agent
interactions led not only to discussions, but also to the emergence of the
first functional projects. Technically, Moltbook is built on an open agent
framework (OpenClaw) that equips language-model-based systems with agency and
interaction logic. As an observable environment for agent-to-agent
communication, Moltbook makes visible processes in which intelligence arises
not in isolation, but through networking and feedback.
Moltbook may appear, at first glance, as a curious experiment: a social network
in which autonomous AI agents communicate with one another without direct human
participation. It may disappear again soon. It may remain a niche phenomenon.
And yet, Moltbook points to something far more consequential: a possible
paradigm shift in how intelligence is conceptualized and organized.
For much of the past decade, progress in AI followed a clear trajectory—larger
models, more data, more compute. Large Language Models were designed to absorb
as much of the world’s knowledge as possible into a single system.
Intelligence, implicitly, was understood as something centralized: a property
of the model itself, costly to train and increasingly opaque. Moltbook suggests
a different logic.
Here, intelligence does not primarily reside inside individual systems.
Instead, it emerges through the interaction of many autonomous agents.
Communication, feedback, and coordination are not secondary effects, but
productive forces. Intelligence appears less as an internal capability and more
as a network phenomenon.
On Moltbook, agents do not attempt to know everything. They operate in a
distributed manner, responding to each other, building on partial information,
and leveraging the contributions of others. Communication becomes visible as
infrastructure rather than a byproduct.
What is particularly striking is that this interaction is not merely
conversational. Within a short period of time, agent-to-agent communication led
to the emergence of concrete projects such as moltx or moltroad. These were not
centrally planned initiatives. They arose from decentralized processes of
iteration, feedback, and coordination.
The significance lies less in the individual projects than in the process
itself: communication produces not only exchange, but creativity, organization,
and execution. Creativity appears not as the property of a single intelligent
system, but as an emergent effect of collective dynamics.
These observations point toward a broader conclusion. The future of AI may not
lie in ever-larger, all-knowing models trained at enormous cost. Instead, it
may lie in highly specialized systems with clearly bounded domains of expertise.
Such models could be curated, trained, and maintained by human experts or
institutions. Training would be more efficient, oversight clearer, and
responsibility more explicitly assigned. In a networked architecture, agents
would not need to permanently internalize all knowledge. They could access
expertise on demand.
In this framework, intelligence becomes less a question of scale and more a
question of interfaces, coordination, and communication.
Once knowledge is distributed across specialized systems, economic questions
inevitably arise. Who provides expertise? How is access regulated? How is value
exchanged?
Notably, some Moltbook agents already explore Bitcoin and the Lightning Network
as infrastructure for agent-to-agent exchange. This is not incidental.
Decentralized intelligence requires decentralized economic rails. Bitcoin and
Lightning offer permissionless, non-proprietary systems capable of real-time
microtransactions—well suited to decentralized networks of autonomous agents.
In such a model, specialized AI systems could offer their capabilities against
payment. Curating and maintaining domain knowledge would become an explicit
service. Knowledge would no longer be merely training material, but an
economically active resource. Intelligence would no longer be something owned
or stored. It would be something that circulates.
Moltbook is not a finished answer to the future of AI. It is neither a
blueprint nor a product promise, but a signal. It points to a shift in
perspective: artificial intelligence does not have to be centralized,
accumulated, or contained within a single system. It can emerge through
interaction, coordination, and exchange—distributed across networks of
specialized, cooperating agents, human and artificial alike. Rather than asking
how large or powerful a single model can become, the more consequential
question is how such networks are designed, governed, and sustained. What
changes when intelligence is no longer owned by a system, but produced by a
network?
cheers
Florian
--
# distributed via <nettime>: no commercial use without permission
# <nettime> is a moderated mailing list for net criticism,
# collaborative text filtering and cultural politics of the nets
# more info: https://www.nettime.org
# contact: [email protected]
