Raw LLM Responses

This is a good read, I say that as somone who works exceptionally deep in the SWE AI space all day every day. One thing that frustrates me in regards to getting involved in the generic AI conversations that you find around here is how whoefully uneducated the public is about how AI is being used in software development at scale and in the most bleeding edge use cases. Without getting into the argument I would point people at the section in this article that describes "multi agent workflows". This is how AI is being leveraged. One thing that the author calls out is that they chose from a couple pre made tools that enabled this ability, they also call out they did not use different models. They chose this option vs creating their own agentic workflows. Organizatons are in fact creating their own multi agentic workflows leveraging MCP and context engineering, specifically they're a creating agents that are bounded to specific contexts and play within their lanes for the most part, for example Architecture mode, planning mode, ideation, implementation, test, integration, etc. where these agents work automously and asynchronously. Memory is also being implemented in a way that gives agents the ability to learn from past iterations and optimize on success. Again not here to argue but I will say using an AI companion chatbot or a place you plug code into and ask for results is like chisseling a wheel out of stone while others are building a rocket to Mars at this point. If you're really interesting in understanding the cutting edge of AI in development I recommend this read as an intro [AI Native Development](https://danielmeppiel.github.io/awesome-ai-native/), full disclosure I'm not the author, but a colleague of mine is.