North American beanbag is back again? Google (GOOG.O) launched the Gemini 3.8 Flash inference model. Wasn't it just half a month ago that Gemini 3.7 Flash was updated, and now 3.8 is here? Speaking of which, ever since Genie switched Openclaw's lightweight task model from Gemini to Deepseek, there haven't been any issues. GOOG has been on a downward trend for the past six months. How about we stop with the big models and follow Meta's lead by selling computing power instead?