Sam Altman at Stanford: AI Should Evolve from Tools to Work Systems
OpenAI CEO Sam Altman stated internally at Stanford that users no longer need to manually write prompts; systems should automatically generate, optimize, and iterate prompts.
The core of his speech is not to teach better prompt techniques, but to transform ChatGPT from a Q&A tool into a work system, achieving task automation and process integration.
This viewpoint reflects a shift in the paradigm of AI applications: moving from "prompt engineering" to the "system building" phase, where the user role shifts from instruction writer to process designer.
Source: Public Information
ABAB AI Insight
Since 2023, Sam Altman has repeatedly emphasized the trend of "AI agency". In his April 2024 Stanford speech, he mentioned that "AI will autonomously execute multi-step tasks", evolving by 2026 into a clear assertion that "systems will autonomously optimize prompts".
On the capital path, OpenAI is transitioning from a "model provider" to a "workflow platform"—launching ChatGPT Work and Tasks features in 2025 to support automated scheduling, information retrieval, and multi-step task execution, forming a three-tier architecture of "model + tools + processes".
This is analogous to the evolution path of Amazon AWS from 2006 to 2010—moving from "providing servers" to "providing managed services" and then to "providing serverless architecture". OpenAI is currently in the "late stage of toolization", with the next step being the "full implementation of agency".
Essentially, this is about technological substitution: AI systems replacing manual prompt writing, as the mechanism for prompt optimization can now be automatically completed through meta-learning, shifting human value from "writing instructions" to "defining goals and acceptance criteria".
ABAB News · Cognitive Laws
Prompts are transitional; systems are the endgame.
Tools help you answer questions; systems help you complete work.
85% of AI's capabilities are not in the dialogue box, but in workflows.