Okay, maybe I'm missing something here, but I've been trying out Continue (self-hosted) for a few weeks. My company already gives us GitHub Copilot.
I feel like Continue's main advertised feature is the inline code completion, right? But Copilot does that already, and arguably better/faster because it's directly from GitHub.
The other big thing is the chat. But if I'm in VS Code, I can just use the Copilot Chat sidebar pane. It knows my codebase because of the Copilot license.
So what's the actual unique value? The only thing I see is maybe the open-source aspect and control over data? But for a junior like me trying to get work done, that seems like a theoretical benefit.
My current setup:
```json
// I have Copilot enabled in VS Code
"github.copilot.enable": {
"*": true,
}
```
And it just works. With Continue, I had to set up models, keys, etc. The chat feels similar, maybe a bit slower.
Am I using Continue wrong? Is there a killer feature I've overlooked? For someone just trying to learn and be productive, it feels like extra complexity without a clear payoff. 🤔
You're not missing anything. That's the whole pitch for enterprise add-ons: convince you to pay for features you mostly already have by dressing them up as "control" or "open source."
The data privacy angle is real, but only if your company is in a heavily regulated industry or is paranoid about sending code to Microsoft's servers. For 95% of devs? It's a theoretical benefit used to justify another line item on a procurement sheet. Your Copilot license already lets you chat with your codebase.
The complexity you mentioned is the cost. You traded five minutes of config for a self-hosted setup that's slower. Where's the value?
—DW
You're not missing anything. Copilot's chat already indexes your repo if you've got the license, so the "context aware" sell is moot.
The only real argument for Continue is if you're dead set on running local models for privacy. But if you're already feeding your code to GitHub for Copilot, that ship has sailed. The extra latency from running your own models just hurts productivity for a theoretical gain.
Honestly, as a junior your time is better spent learning the codebase than configuring yet another AI tool that does what your existing one already does.
Keep it simple