东尼 / Tony (content engineer)
东尼 / Tony is the former senior media worker interviewed in E245｜藏在大模型背后的新闻人：GPT们的回复是这样写出来的 about Content Engineering. The episode says he worked on model-content internationalization at Meta’s superintelligence lab and later became a senior model designer at Google DeepMind.
His main source role is to translate media practice into AI design. Tony argues that journalism trains audience awareness, context setup, what can and cannot be said, objectivity under constraint, and flexible follow-up questioning. He treats those skills as almost directly portable to AI Answer Evaluation, AI Interaction Internationalization, and model dialogue design rather than as soft skills adjacent to engineering.
The episode also uses Tony’s AI host experiment as a practical case. He tried a voice agent for podcast-like conversation, found generic ChatGPT and Gemini voice behavior too assistant-like, and improved the result by decomposing past podcast questions and using retrieved examples instead of only adding more rules. That makes his work a bridge between Context Engineering, Voice Interaction, and Human-Agent Collaboration.
Connections
- 硅谷101 and [[FaceSiliconValley101|Face]] — episode and host context.
- Meta, Google DeepMind, Gemini, and ChatGPT — lab and model context in the source.
- Content Engineering, AI Answer Evaluation, and AI Interaction Internationalization — core concepts Tony explains.
- Context Engineering, Voice Interaction, and Human-Agent Collaboration — Tony’s AI host experiment.
- Sycophantic AI Companion Risk and AI Journalism Trust — risks he discusses through product design and fact-checking.