QCon San Francisco 2026: The Harness Is the New Operating System
At QCon San Francisco 2026, Mad Labs founder and CEO Manju Rajashekhar opened the conference arguing that the harness around an AI model, rather than the model itself, decides what an agent can do and how safely it does it. The keynote mapped the parts of a harness, including context, tools, state, permissions and feedback loops, onto jobs that operating systems already perform. By Andrew…
QCon San Francisco 2026 featured a keynote titled "The Harness Is the New Operating System," delivered by Manju Rajashekhar, founder and CEO of Mad Labs. Rajashekhar argued that the harness surrounding an AI model, rather than the model itself, determines an agent's capabilities and safety. He likened the harness to various components of an operating system, including context, interface, execution, and feedback mechanisms.
The harness is responsible for providing tasks, context, tools, state, permissions, and defining what constitutes success or failure, while the AI model itself generates behavior at runtime. Rajashekhar emphasized that the harness is essential for making AI agents dependable and reliable software, highlighting the importance of isolating errors and ensuring that each iteration of an agent's loop counts.
Brief written by urgent.news; may contain errors — check the original.
Source: infoq.com