Open-source projects chained LLMs with tools and memory for multi-step goals—high hype, mixed reliability.
In spring 2023, AutoGPT, BabyAGI, and similar projects viralized agents that decompose goals, call tools, and replan.
In practice they failed via loops, cost, and plan hallucinations. Still, they educated the market on tool use, memory, and trajectory evals.
The second wave (2024–2026) industrialized agents with better sandboxes, tracing, and reasoning models—but the founding myth was 2023.
Full-duplex realtime audio APIs enabled low-latency call centers and spoken assistants with tool use.
Enterprises deployed agents with real tools—CRM, code, browser—under tracing, permissions, and continuous eval.
Labs showed models that click, type, and navigate UIs like humans, extending automation without APIs.