AIEWF 2026: Experts Debate Human Agency, Autoresearch, and the Limits of Full Automation
English summary
At AIEWF, Introspection co-founder Roland Gavrilescu defined autoresearch as an outer loop where agents study and maintain the primary system. Anthropic’s Thariq Shihipar described how Claude Code is “grown, not developed” through continuous user-driven discovery. Addy Osmani argued that the inner execution loop belongs to agents (capability), but the outer loop of goal-setting and judgment must stay human (agency). Paul Bakaus launched Impeccable, a design tool that refuses one-shot solutions, requiring human involvement for the final 20% to add taste and ownership. Panels on generative media and agentic sites reinforced the need for human sensitivity, creative direction, and brand stewardship even as models advance.
Chinese summary
在AIEWF上，Introspection联合创始人Roland Gavrilescu将自动研究定义为由代理维持系统的外部循环。Anthropic的Thariq Shihipar说Claude Code是“长出来的，不是开发出来的”，通过持续的用户驱动发现成长。Addy Osmani认为内部执行循环交给代理（能力），但目标设定和判断的外部循环必须由人类保留（能动性）。Paul Bakaus推出了设计工具Impeccable，拒绝一键解决方案，要求人类参与最后的20%以注入品味和所有权。关于生成媒体和代理网站的讨论均强调，即使模型不断进步，仍需人类的敏感性、创意方向和品牌管理。
Key points
Roland Gavrilescu defined autoresearch as an “outer loop” where agents study and maintain the primary system.
Roland Gavrilescu将自动研究定义为由代理研究和维持主系统的“外部循环”。
Anthropic's Thariq Shihipar stated that Claude Code models are “grown, not developed” through continuous learning during use.
Anthropic的Thariq Shihipar表示，Claude Code模型是“长出来的，不是开发出来的”，通过使用中的持续学习完成。
Addy Osmani distinguished the inner loop (agent capabilities) from the outer loop (human agency), arguing the latter must stay with engineers.
Addy Osmani区分了内部循环（代理能力）和外部循环（人类能动性），主张后者必须留在工程师手中。
Paul Bakaus introduced Impeccable, a design tool with no auto mode, forcing humans to steer the final 20% for taste and authorship.
Paul Bakaus推出了设计工具Impeccable，没有自动模式，强制人类引导最后20%以体现品味和作者身份。
Generative media panelists Nicole Brichtova and Shane Gu emphasized that human sensitivity and cultivated expertise remain essential to avoid generic AI output.
生成媒体讨论者Nicole Brichtova和Shane Gu强调，人类的敏感性和培养的专业知识对于避免通用AI产出仍然至关重要。
Adobe's Carlos Sanchez demonstrated agentic sites but cautioned that full AI generation can violate brand guidelines without human oversight.
Adobe的Carlos Sanchez展示了代理网站，但警告称，没有人类监督的全AI生成可能偏离品牌准则。