← Back to context

Comment by stavros

17 days ago

OpenCode and Pi do those things as well, and without a whole annoying IDE bundled in.

OpenCode is miserable from a security perspective. Well clarification the plans they offer where they bundled in free models that train on your use. You are then left to use an OpenRouter which I find pretty flaky for at least the leading Chinese models.

  • I doubt most people use OpenCode coding plans, nor do they use OpenRouter. I use subscription plans from ChatGPT, z.ai, MiniMax & Xiaomi with OpenCode. It handles authentication with all of them seamlessly. I switch between models based on task/subtask and based on usage limits. You can get the most value out of a lot of these plans at their second-tier and they are often switching in value relative to each other, so it makes sense to arbitrage them like this.

    Most of that switching is automated (oh-my-openagent - defaults sub-tasks to different roles, so for example I use MiniMax for explorer tasks and GPT 5.5 for deep design & review tasks, and GLM 5.2 for general orchestrator & most coding). If I hit usage limits it switches to a backup for that task. I'm not sure Cursor authenticates with all the subscription coding plans from all those companies - but if it does it can't be doing it any better.

    I run it in a sandbox and its not phoning home.

    • Which is cool but I think it’s an important callout because it’s shady how they do it in my opinion.

  • I just use my ChatGPT subscription with it. Not sure what you mean about security.

    • “Well clarification the plans they offer where they bundled in free models that train on your use.”

      Just what I said. They offer paid plans through their tool. Said paid plans are kind of a dark pattern where it’s not immediately obvious the models are training on your data. The harness is fine but that kind of business turns me off and I am usually pretty neutral about those sorts of things.