CodeAct and Manus

CodeAct, one of the frameworks / AI Agent implementations that powers startups like Manus, offers an interesting alternative to function calling. Rather than having the LLM output structured outputs used to call a function, CodeAct proposes that you should let the LLM actually code, act, and revise based on what it observes. On benchmarks, they're able to achieve better performance using this approach. Compared to function calling, CodeAct allows for more flexible solutions. ---- Hi, I’m Diana 👋 I’m a 4x first enterprise software PM, co-founded my own startup, and worked in venture capital! I double click into important trends in tech, startups, and AI without the hype. Follow for more! ---- #codeact #llm #aiagents #manus