Hacker News Reader

Comment by mdasen | Hacker News Reader

Is Claude Code better than the Gemini CLI? I've been using the Gemini CLI with Gemini 2.5 Pro and haven't been impressed. Maybe these LLMs aren't as good with Rust codebases? I'm guessing there are a lot more people looking to use these tools with JS and Python.

View on HN

Gemini CLI is terrible. I've had it:

- Repeat more than 20 times the same response to my prompt rejecting its proposed changes; I just kept prompting to see how far it would go before doing something different. Claude Code would quickly guess there is something wrong and try something else or ask what I'm getting at

- Continually refer to outdated versions of files, even after I've told to re-read the files

- Refer to files in a different session on a different machine that have no relevance to what I'm currently doing, presumably simply because I logged in with the same account.

- Randomly crash or enter infinite loops, sometimes soon after starting

- Refuse to read files in a sibling or parent folder

- Fail to understand simple request.

- Propose empty changes

Claude Code is just far better. I only use Gemini CLI for the simplest of tasks

You can make Gemini CLI much better by making it behave more like Claude Code. Claude Code has some lovely prompt engineering at the system and subsystem level that can be replicated with Gemini CLI. I’m having great results already. I am still perfecting process and prompts to be a fully agentic system that can do well on benchmarks but more importantly do the right work with steerability, which was an absolute pain with Gemini CLI out-of-the-box. If you are interested, I can publish some of the basics now and then I can keep you posted as I develop it into a more robust system. Just email me at randycarlton@gmail.com with the subject: SaaS.bot (where this work will likely reside).

I've been using both on a Rust codebase and have found both work fairly well. Claude code is definitely more capable than Gemini. What difficulties have you had?

The biggest pain point I've had is that both tools will try to guess the API of a crate instead of referencing the documentation. I've tried adding an MCP for this but have had mixed results.

https://github.com/d6e/cratedocs-mcp

It might be that we have multiple features in our codebase and Gemini seems to struggle understanding that it needs to be aware of #[cfg(feature = "x")] and also that if it's trying to run things, it might need to specify the feature.

And yes, when they guess APIs, it's highly annoying.

i've tried codex, cursor, and a few other agentic tools and nothing compares to claude code when it comes to UX. The other service's models are quickly catching up to claude, but the claude code ux is just magical to me. i havent used it with rust personally. like you suggested would be the average user, i've mostly stuck with js and python.

I don't know if it's Gemini CLI or Gemini 2.5 Pro, but the combination is not even comparable to Claude Code with Sonnet. I was starting with these agent tools several weeks ago, so it was very tempting to use Gemini, instead of paying for Claude Pro, but the difference is huge. In my experience, Gemini was very quick to get stuck in debugging loop, fixing something "one last time" over and over again. Or it got into writing code, despite my explicitly saying not to do so. I'm still trying to figure out if I could use Gemini for something, but every time I try it, I regret it. Claude Code with GLM-4.5 is a good alternative to paying for Claude Pro, it's not as good as Sonnet, but close.

I guess what seems surprising to me is that Gemini 2.5 Pro scores well above Claude Sonnet on Aider's leaderboard, even beating Claude Opus 4.

I have been kinda wondering if there's something that just isn't as good between the CLI and model because the Gemini CLI has been a mostly frustrating experience - and it's kept me from wanting to pay for Claude because I don't want to pay money for the same frustrating experience. But maybe I should try Claude and see.

https://aider.chat/docs/leaderboards/

I was once a heavy user of Cursor with Gemini 2.5 Pro as a model, then a Claude Code convert. Occasionally I try out Gemini CLI and somehow it fails to impress, even as Cursor + Gemini still works well. I think it's something about the limited feature set and system prompt.

I have found Claude code to be significantly better, both in how good the model ends up being and in how polished it is. To the point that I do not drop down to Gemini CLI when I reach my Claude usage limit.

It is much better but only because Sonnet 4 is better at handling more complexity and being very code at writing code.

Claude Code is insanely good with Rust.

Reply on news.ycombinator.com