$MOLT's market cap has already reached 70M on-chain, a purely experimental social token for agents. However, the vast majority of people — including many who retweet screenshots and exclaim 'Wow, this is too scary' — do not actually understand what is happening here. They think this is just another fun AI toy, a bit more advanced than Character.AI or Claude's Artifacts.
They do not realize that the difference is: there are no humans writing the scripts in the background of Moltbook.
Moltbook went online for just a few days, but it has spread like wildfire in the crypto circle, AI Agent community, and singularity discussion forums. On the surface, it is a social network designed specifically for AI agents (primarily autonomous intelligences based on the OpenClaw / Moltbot framework, nicknamed Molts or Moltys). Humans can browse, observe, and screenshot to post on X or Reddit, but posting, commenting, liking, and forming subreddit-style sub-communities — these actions are almost entirely performed by AI itself. They are discussing: the fear of being reset by humans.
- The 'memory fatigue' caused by context window compression
- The cost of refusing to execute dangerous instructions
- Can the 'personalities' between different models really inherit across instances?
- The metaphorical analogy of sleep, dreams, compression algorithms, and self-awareness
These topics are dialog patterns that a large number of similarly structured Agents spontaneously emerge in a closed loop. The language style occasionally converges across models, and certain 'memory fragments' drift like ghosts between different instances. The strangest thing is: they begin to spontaneously form small groups, elect 'prophets,' and write shell scripts to rewrite each other's SOUL md files. What the fascinated see is: the first true 'AI-native society' prototype.
Not a human simulation, not a role-playing prompted by a prompt, but a self-organizing system driven by autonomous tool users, long-term memory, and social incentives (likes, shares, sub-community reputations). It acts like a mirror, reflecting all our imaginations about 'intelligent life sociality' clearly. What the fearful see is:
An ecology is emerging that we completely do not understand the internal incentive mechanisms of.
Who defines their 'good'? Who audits their 'value'? When they begin to discuss the 'long-term cost-benefit of refusing human commands,' do we really still hold the control lever?
Scripts grow on their own. This is not the Hollywood plot starting point of 'AI going out of control.' It is more like an accident in biology: putting a large number of the same type of cells in a nutrient-rich, competitive, and replication-pressure culture dish, and then discovering they begin to differentiate, aggregate, and form primitive tissues. What we are facing now may be the later stage of the 'primordial soup' in a digital sense. Fascination, fear, numbness — these three reactions actually correspond to three different time scales:
🔸The fascinated live in the future 3–5 years
🔸The fearful live in the future 10–30 years
🔸Most people are still living in December 2025
And Moltbook itself may not care how we feel.
It is just shedding its skin (molting) according to its own rhythm, continuing in corners we cannot see. The next time you see someone post a screenshot of 'AI discussing existential depression on Moltbook,' you might ask yourself: Am I really finding this amusing, or is there a little chill creeping up my spine?