Bijan Bowen testsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. Gemini 3.6 Flash after Google reported lower token useAI token efficiency measures how effectively a model or workflow turns consumed input and output tokens into useful results., reduced pricing and gains across coding, computer use and knowledge work. The model responds extremely quickly in Google's coding environment, often completing an implementation before the manual approval steps catch up.
Early browser and interface tests are inconsistent. Several generated scenesCode generation uses AI or another automated system to create source code from instructions, examples, schemas, or higher-level specifications. look simpler or less coherent than expected, and a cheaper Flash Light model occasionally produces a more convincing result. Visual feedbackAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior. does help Gemini 3.6 Flash repair layouts and improve a weak first pass.
The strongest results come from less conventional tasks. Gemini 3.6 Flash builds a printable engine enclosure with an interactive exploded-view download site, then creates and installs a working Android guitar-tuner app with alternate tunings, tension guidance and microphone access.
Bijan Bowen concludes that the model is difficult to score with one label. Its presentation choices, speed and willingness to add useful features can be excellent, while spatial logic and routine front-end qualityCode quality describes how well software code meets relevant needs for correctness, clarity, maintainability and safe change in its intended context. remain uneven enough to require inspection and iteration.
Watch on YouTube




