On September 28, 2026, Google presented four projects built with Gemini 3.8 Flash. The model handles different kinds of input and is intended for multistep reasoning and software tasks, according to Google. These are examples from individual builders. They show possible workflows, but the post does not establish that every user will obtain identical results from the same model. Source
The four projects
One builder mapped live paths for satellites, a space station, and rockets. Another animated a Japanese wave pattern. A dinosaur skeleton project used a four-stage prompt with constraints and accuracy checks. An interactive automatic-transmission model included ten camera views, multiple display modes, labels, and a panel explaining its parts. The examples span both visual work and interactive technical explanations. Source
Where to experiment
Google points to AI Studio and Antigravity as tools for trying Gemini 3.8 Flash. AI Studio is a development environment for testing model inputs and outputs. A reader can inspect the linked projects in the original post to see which task, prompts, and interface each builder used before attempting a similar experiment. Source