Hacker News Reader

Comment by gorjusborg | Hacker News Reader

Flip your thinking around for a second and consider why an IDE is required for an agent that codes for you?

The IDE/editor is for me, the agent doesn't need it. That also means I am not forced to used whatever imperfect forked IDE the agent is implemented against.

View on HN

> why an IDE is required for an agent that codes for you

Because the agents aren't yet good enough for a hands off experience. You have to continuously monitor what it does if you want a passable code base.

Sure, but monitoring, reviewing and steering does not really require modern IDEs in their current form. Also, I'm sure agents can benefit from parts of IDE functionality (navigation, static analysis, integration with build tools, codebase indexing, ...), but they sure don't need the UI. And without UI those parts can become simpler, more composable and more portable (being compatible with multiple agent tools). IMO another way to think about CLI agentic coding tools as of new form of IDEs.

As was already mentioned elsewhere, Emacs + Magit to monitor incoming changes is a great combo.

yes i am rejigging my whole vim setup

Following are now stars of my workflow

* Git plugins - Diffview, gitsigns, fugitive

* Claude Code plugin / Terminals with claude code

* Neovim sessions

* Git worktrees

Editing focused workflows have taken an backseat

* LSP

* Vim motion and editing efficiency

* File navigation

* Layouts

why even use vim at this point? the LLM ecosystem there is decent, but definitely less polished than using a modern IDE

Vim motions are nice.

I don't really need an IDE, but I do need a great code review interface.

As someone who hasn’t used Claude Code yet, can’t you configure it somehow to use a different tool of your liking, or it has to be in the cli?

I end up using the VCS tooling (lazygit for me), but coding agents really need to be integrated with this review environment. We need an extra step where the agent will group its changes into logical units (database models in one commit, types in another, business logic in another, tests in another), rather than having to review per-file.

Programming has changed from writing code to reviewing/QAing and reprompting, but the tooling hasn't yet caught up with that workflow. We need Gerrit for coding agents, basically.

I just merge the change and review the diff. If it’s wrong I either revert or ask Claude to fix it.

I use lazygit for that. But any diff tool you like will work.

Whenever I have to take the wheel myself the AI tab completion makes it much smoother so I am kind of addicted to that. Semi-automatic mode.

I would much rather use IntelliJ so perhaps my habits will change at some point, but right now I am stuck with Cursor/vscode for the tab completion.

Reply on news.ycombinator.com