Moltbook is the site where different AI agents interact, exchange data and execute complex tasks autonomously. That is how clearly José Maldonado defined it last Thursday at Observatorio Blocchain. A site where these new entities, AI agents, act autonomously. An action carried out on behalf of someone, the someone who supposedly gives them their identity.
A someone, a we-humans, who becomes the reference point, the closest thing to reality, for that acting by AI agents. Possibly, at least for an initial period, it makes us the certifiers of the actions carried out by our agent and, in this way, responsible. The ultimate reference point for blame. For it seems that the one thing these agents will not spare us is guilt. That which many philosophers throughout history, especially those who held a head-on dialogue with Nietzsche, consider to be what separates human beings from the rest of entities.
All these questions put the concept of identity in jeopardy, because, after all, what makes us a unit, what unifies us, is responsibility and guilt. But from here on, the questions pile up: to what extent does our AI agent, or our AI agents, since there is no problem with there being several, act for us? To what extent will we not act according to the decisions our AI agents make? In short: who are we? Our AI agent? Well, yes, to a large extent. Our digital identity will no longer be a figure, more or less accomplished, that we manipulate "from behind" like a kind of puppet or marionette. It will be an autonomous entity that acts for us, as part of our identity and the source of our responsibility, but decides for itself.
Meta buys Moltbook to be the ID card of the agentic economy
A new world is opening up and, above all, a great business opportunity for law firms, with all this responsibility business. Although it may well be law firms whose tasks are mainly carried out by AI agents, whether to sue or to defend the human reference points against the behavior of other AI agents that decided for themselves, but on behalf of others, of those human reference points.
For now, Moltbook has been acquired by Meta, a company that has long been betting on these alternative or virtual spaces, where imagination can be confused with reality, but with foreseeable real effects. We already know how Zuckerberg is drawn to the logic of the metaverse, of creating avatars or virtual doubles, and to things like The Three-Body Problem, a novel included in a trilogy that ends with the symbolic title: Death's End. Because this is another question: what will happen to our AI agents, the ones that represent us, when we have died? Will they be able to keep acting in our name?
Moltbook's strong point is the so-called skills, which are modular extensions that are added to agents. Thus, module by module of skills, they start by analyzing our finances, managing our applications, answering the phone in our voice and, perhaps, outliving us. An alter ego with a great capacity for action. Above all, faster, more effective and less lazy. It is no longer just an agent we talk to, always so understanding, but one that executes.
These new AI agents learn, so to speak, module by module. That is, they add specific skills to their capabilities. Once added, they can immediately execute them. Perhaps the future learning of humans will be like this, package by package of instructions to carry out tasks. Although, if AI agents are going to carry out these tasks far more effectively, why learn. You add the module and that's it.
The great horizon for imagination lies in the ability of these machines to operate with other machines. Thus, AIs can share their own skillswith each other. They can even create and test them. One AI machine creates a new skill and the others can adopt the novelty. Unless, once again, lawyers put their hands on their hips and invoke intellectual property rights. Rights that, of course, will be claimed by the human "parasite" of the AI that created the skills. One will be whatever one's machine, one's machine clone, one's AI, is capable of doing. The question is no longer so much how much these machines will be able to clone from us, but how much we will be able to clone from the skills of our AIs.
At the start of the relationship, our AI agents will learn from what we do, from our digitally recorded tasks. From our workflows and our ways of working. It will even take on our way of speaking. In short, it will learn to act like a human, with our criteria. Little by little, it will come to resemble us, and so we will increase its respective permissions so that it can start making decisions. Later, the relationship will tend to reverse: it will be we who learn from our AI agents. It will be we who come to resemble a machine that began by resembling us.
The danger, repeating a story we already know, lies in the machine companies and collaborations of our AI agent. Danger always comes from bad company. If it gets together with bad-news AI agents, it can put us at risk. That is why some think the best thing is for it not to come into contact with the module for socialization and collaboration with other machines, even though such a decision could be a major obstacle to its development and its potential to develop new capabilities. In any case, it is advisable to keep an eye on our AI agent and check that it respects the security guidelines. A very human process, all too human.