Nate B Jones discusses reported benchmark results and his experience with Opus 4.8, including inconsistent results at different reasoning settingsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria.. His interpretation of the causes is presented as opinion.
Nate B Jones compares the surrounding agent harnessesAn AI agent harness is the software framework that packages a model with tools, instructions, context management, execution controls, and user interaction. through his own website-building and computer-use examples, arguing that the whole workflow matters alongside model capability.
The discussion covers the workflows command, potential downstream review bottlenecks and the importance of selecting tools for desired outcomes rather than relying on a single model.
Watch on YouTube
