Nate B Jones uses a 514-piece LEGO version of his logo to illustrate Opus 5.5's ability to produce a coordinated model, assembly animation, instructions and parts lists. He reports that the task used a small share of his subscription allowance, but distinguishes that subscription experience from an estimated API billCost per completed task measures the total AI, tool, infrastructure, retry, and repair expense for each verified useful outcome. and warns that other tasks can have very different costs.
Nate B Jones emphasizes steerability in writing and code-driven visual work. Useful revisions should preserve the user's meaning, uncertainty and design decisionsAI instruction following is a model's ability to understand and reliably comply with a user's valid requirements and constraints. while changing the requested details. He describes better results with long-running engineering tasks when the permitted scope, stopping conditions and definition of done are explicit.
Nate B Jones connects these improvements to user feedback and AI-assisted development inside model laboratories, without claiming access to their full internal development histories. His practical recommendation is to repeat a representative assignment from the same files and promptEvaluation measures how well an AI system performs against defined tasks, criteria and failure conditions using repeatable evidence., track unsuccessful attempts and human corrections, and compare usable finished results across model releases.
Watch on YouTube
