Hermes is especially powerful for DevOps and the work surrounding your code. I use it to manage deployments, servers, configurations, migrations, logs, dependencies, and infrastructure. It can also make code changes, debug problems, modify configuration files, and work across multiple machines through tools like SSH.

For new projects, I recommend starting with planning rather than immediately generating code. Skills such as Matt Pocock’s Wayfinder and interview-style planning skills can question you about what you’re trying to build, uncover missing requirements, and turn an idea into a specification, roadmap, and implementation plan. Giving the coding agent a strong plan usually produces much better results than simply asking it to start building.

For small applications, scripts, utilities, prototypes, and one-off artifacts, Hermes is incredibly fast. You can simply ask for what you need in the conversation you’re already having, and it can create, test, debug, and iterate on it without switching environments.

For long-running software projects with lots of continuous work inside a single repository, I still tend to use a dedicated coding terminal. Coding terminals provide a focused environment for extended development sessions, while Hermes can work alongside them—handling research, infrastructure, debugging, deployments, and everything around the project.