Ask HN: Am I using AI wrong for code?

The author expresses a concern about not fully utilizing AI tools for coding, despite having tried options like Copilot and others that did not integrate well with their workflow. Currently, they primarily use Claude for brainstorming architecture, tool recommendations, and obtaining small code snippets, which has improved their productivity and confidence in tackling unfamiliar programming languages and stacks. However, they feel they are missing out on the potential of AI, especially with advancements in large language models (LLMs), retrieval-augmented generation (RAG), and fine-tuning, which could assist in more substantial coding tasks with supervision. The author has observed reports of AI successfully aiding in significant projects like refactoring and feature development. They seek recommendations for tools that can suggest and iterate on edits across multiple files, incorporate context from selected documents or codebases, and allow for branching conversations to facilitate learning without disrupting the main discussion. Although they have begun developing a command-line interface (CLI) for this purpose, they prefer to find an existing solution.