Hacker News Reader

Comment by xnx | Hacker News Reader

The next realization will be that Claude isn't clearly(/any?) better than Google's coding agents.

I think Gemini 3.0 the model is smarter than Opus 4.5, but Claude Code still gives better results in practice than Gemini CLI. I assume this is because the model is only half the battle, and the rest is how good your harness and integration tooling are. But that also doesn't seem like a very deep moat, or something Google can't catch up on with focused attention, and I suspect by this time next year, or maybe even six months from now, they'll be about the same.

> But that also doesn't seem like a very deep moat, or something Google can't catch up on with focused attention, and I suspect by this time next year, or maybe even six months from now, they'll be about the same.

The harnessing in Google's agentic IDE (Antigravity) is pretty great - the output quality is indistinguishable between Opus 4.5 and Gemini 3 for my use cases[1]

1. I tend to give detailed requirements for small-to-medium sized tasks (T-shirt sizing). YMMV on larger, less detailed tasks.

Claude is cranked to the max for coding and specifically agentic coding and even more specifically agentic coding using Claude Code. It's like the macbook of coding LLMs.

Claude Code + Opus 4.5 is an order of magnitude better than Gemini CLI + Gemini 3 Pro (at least, last time I tried it).

I don't know how much secret sauce is in CC vs the underlying model, but I would need a lot of convincing to even bother with Gemini CLI again.

That hasn’t been my experience. I agree Opus has the edge but it’s not by that much and I still sometimes get better results from Gemini, especially when debugging issues.

Claude Code is much better than Gemini CLI though.

Reply on news.ycombinator.com