Within three to eighteen months, AI models will perform end-to-end tasks for users, which will fundamentally change how people view AI.

The claim has two parts, both due by October 2025: (a) AI models that carry out end-to-end, multi-step tasks for users (agentic products that complete a task rather than just answering) become publicly available; (b) this fundamentally changes how people view AI. TRUE: by 31 October 2025, at least one major provider has released generally available agentic products that complete end-to-end tasks for users (for example computer-use or browser agents, or autonomous coding agents), and representative surveys or broad, credible coverage document a marked shift in public or professional perception of AI, from chatbot to agent or worker. MOSTLY TRUE: both parts hold by the deadline, but the products are limited (preview or paid tiers only) or the perception shift is documented mainly among professional or technical users. MIXED: part (a) clearly happened by the deadline, but there is no clear evidence of a fundamental perception shift, or the evidence on it is contested. MOSTLY FALSE: only limited research previews of end-to-end agents exist by the deadline, or they arrive after October 2025, and there is no perception shift. FALSE: no publicly usable end-to-end task agents exist by October 2025.

Set before the research began.

1. E1 Introducing Operator OpenAI · 23 January 2025

   “Operator can be asked to handle a wide variety of repetitive browser tasks such as filling out forms, ordering groceries, and even creating memes.”

2. E2 Introducing Operator OpenAI · 23 January 2025

   “Operator is now fully integrated into ChatGPT as ChatGPT agent.”

E1 and E2 show that OpenAI announced Operator, a browser task agent, in January 2025 and later integrated it into ChatGPT. They do not show whether it was generally available or only a limited preview, or when it was integrated, so they cannot tell whether part (a) held by October 2025. No evidence covers a change in how people view AI.