[nextaccountic]

In the front page there's a project that attempts to reduce tje boilerplate of mcp output in claude code

Eventually I hope that models themselves become smarter and don't save the whole 54k tokens in their context window