Gemini vs Doubao: Perception, Competition, and Adoption

A social media post by @BTCdayu claims that Google’s Gemini is “the U.S. version of Doubao” and describes it as “good-looking but not useful,” implying the product appears impressive but performs poorly in practice. No supporting details, examples, benchmarks, or dates are provided beyond the title, and there is no accompanying article body to clarify which Gemini features are being criticized or how it compares to Doubao (ByteDance’s AI assistant). As presented, the item reflects a user opinion rather than a reported product update, test result, or official statement. The limited information makes it impossible to assess the accuracy of the comparison or the specific shortcomings being alleged.

任何以 AI 競賽已經結束某某公司遙遙領先另一家公司已經沒救為基礎的論述，基本上都可以跳過不看，沒有參考價值。AI 競賽現在也只是三局上半，任何領先都可能被逆轉。 來看看過去幾個月的例子就知道了，Google 在去年底 Gemini 3.0 推出被吹捧成終局贏家，結果二月份就被 Anthropic Claude 4.6 在真實工作環境全面輾壓。Anthropic 靠 Claude Opus 4.6 在 agentic workflow 被認為遙遙領先了三個月，好像 AI 賽局已定 Anthropic 成唯一贏家，但四月底 OpenAI 靠 GPT-5.5 就達成反超。這樣的彼此超車，過去兩年已經演了好幾輪，估計未來幾年也還是會一樣。 所以，真的不要因為最近某家 AI Lab 的東西比較強，就把世界觀直接設定成比賽結束了。我認為無論是 Meta 或者是 SpaceXAI，絕對都還有戲。甚至微軟的 MAI 與亞馬遜的 Nova 未來也會進入競爭區。

Simon Willison released llm-gemini 0.31, an LLM plugin that provides access to Google’s Gemini family of models; the announcement notes that gemini-3.1-flash-lite has graduated from preview to generally available in this update. Willison links back to his March write-up of the Gemini 3.1 Flash-Lite preview and suggests the non-preview model behavior remains unchanged. The post is a brief release note aimed at developers and practitioners integrating Gemini models via the llm-gemini plugin, signaling stability and production-readiness for gemini-3.1-flash-lite. This matters to developers building AI features or tooling that depend on Google’s latest efficient multimodal models.