Saying LLMS are not good at x or y, is akin to saying a brain is useless without a body. Which is obvious. The success of agentic coding solutions depends on not just the model but also the system that the developers built around the model. And the companies that will succeed in this area are going to be the companies that focus on building sophisticated and capable systems that utilize said models. We are still in very early days where most organizations are only coming to terms with this realization... Only a few of them fully utilize this concept to the fullest, Claude code being the best example. The Claude models are specifically trained for tool calling and other capabilities and the Claude code cli compliments and takes advantage of those capabilities to the fullest, things like context management among other capabilities are extremely important ...