Nvidia announced NOOA (Object-Oriented Agents) last week, signaling a shift in agent development where the harness around a model may matter as much as the model itself. Nvidia is contributing NOOA to the Open Secure AI Alliance, an industry group Nvidia formed last week to build and share open-source and open-weight tools for AI development. NOOA is founded on the idea that an agent is a single Python class, with methods as capabilities, fields as state, docstrings as prompts, and type annotations as enforced contracts. A standard Python method whose body is an ellipsis (…) is completed at runtime by an LLM-driven loop, while methods with normal bodies run as ordinary, deterministic Python. By combining capabilities, state, and prompts in one class, Nvidia aims to address fragmentation in agent development. Adnan Masood, Ph.D., chief AI architect at UST, noted that the harness is currently scattered by design, with prompts in Jinja templates, tool definitions in JSON schemas, callbacks in Python, and workflows drawn as graphs. Nvidia’s approach makes agent development more like traditional software development, simplifying testing and tracing. Karthik Karunanithi, solution architect at IBM, acknowledged the improvement but questioned whether it may become harder to distinguish deterministic code from probabilistic behavior, since both share the same signature, indentation, and docstring. Siddhartha Saxena, co-founder of Thine and Merlin AI, credited Nvidia for using