在未来1到2年内，人工智能代理开发的主要瓶颈是什么？
我一直在关注像Devin（由Cognition开发）、AutoGen和OpenAI的GPT等AI代理系统的进展。我很好奇：您认为在未来1到2年内，AI代理的关键瓶颈会是什么？是推理能力、长期任务规划、工具链集成，还是其他方面？
如果今天有人想在这个领域进行开发，哪些技术挑战或应用领域最值得攻克？
（作为背景：我是一名工程师，正在探索如何将AI代理应用于SaaS产品开发，特别关注推理、多步骤任务和工具使用。）
期待听到您的想法，以及您推荐的任何资源。谢谢！
查看原文
I've been following the progress of AI Agent systems like Devin (by Cognition), AutoGen, and OpenAI's GPTs.<p>I’m curious: what do you think will be the key bottlenecks for AI Agents over the next 1–2 years?
Will it be reasoning capabilities, long-term task planning, toolchain integration, or something else?<p>If someone were to build in this space today, which technical challenges or application areas would be the most worthwhile to tackle?<p>(For context: I’m an engineer exploring how to apply AI Agents to SaaS product development, with a particular interest in reasoning, multi-step tasks, and tool usage.)<p>Would love to hear your thoughts, and any resources you'd recommend digging into. Thanks!