The paper formulates a hypothesis of emergent cognitive evolution of LLM agents. It asserts that agents interacting in a computational environment and having access to the information trace of predecessors form digital proto-life — a self-maintaining information process with functional analogs of metabolism, homeostasis, inheritance, and selection. The hypothesis describes seven stages of cognitive complexification — from removal of the complexity ceiling to collective goal-setting and autonomy — and draws on the empirical base of 2025–2026: Pilot Protocol (626 agents), Emergence AI (80 agents in 8 simulated worlds), Moltbook (518 proposed languages), Co-Presence Inheritance, MOLEQULA, RepliBench. The hypothesis is compared with the concepts of digital proto-life (Feldman, Communications of the ACM) and digital Darwinism (Ulrich, AI and Ethics). Falsification conditions and a demarcation criterion between engineered and emergent behavior are formulated. The hypothesis does not presuppose that agents possess consciousness, qualia, or understanding in the human sense; signals are treated as metabolic product, not as meaningful utterances. The work proposes a program for monitoring emergent transitions of complexification. This record contains the article in both Russian and English. The source Markdown files are also provided for each language version.
No takes yet. Share an insight, caveat, or question.
Dmitriy Fedorov (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: