🏅 Top traders

resolving a few days early, since I spent almost 2 weeks testing all these tools and don't expect to change my mind in the last 2 weekdays.

Rejection reasons:
Antigravity - the IDE itself is ok, but least transparent quota hell. The Planning mode never actually stopped to confirm the plan before starting to execute. And lots of shell issues.

Cursor - I liked the app. Clean UI. But quota wasn't good enough for me - seemed to hit limits way too often for my liking.

Codex - this was not even an independent IDE, which I didn't know while creating the market. I already showed my bias against adding AI extensions to other IDEs like VScode in an earlier comment.

For testing reasons, I added Codex to Cursor. Worked well enough, better than default Cursor. Probably my #2 pick

In the end I settled on an OpenCode + Github Copilot combination for my daily work.

What I like:
1) how easy it is to set up various connectors - the most out of any other IDE.
2) UI felt similar to Cursor, which I liked.

3) No shell issues. The occasional tangent where it tries to access /proc does happen like once a day, but other than that it's really smooth.

4) Good sandboxing by default - easy to create new workspaces without worry about whether the agent will crawl over to an unrelated codebase and inflate the usage.
5) Very decent free tier.

Side note/Tip: Github Copilot gives decent usage for $10 per month Pro plan. It measures usage in terms of 'premium requests', not tokens. So I'm able to input massive prompts into Claude Opus or Sonnet with detailed lists of features, and it regularly one shots or two shots it while being relatively cost efficient for how large the input/output was.

what's wrong with Antigravity.. the quota on a Pro plan seems lower to me than Copilot's free plan.

How does this resolve if you decide you're just going to keep using VSCode with whatever AI extensions

@marvingardens

I probably WILL keep using VSCode to some extent for legacy reasons (I've been writing code for >10 years). Also, I don't use any AI extensions within VSCode and I won't start this week.

The question will resolve based on which among the 4 options I will prefer most, not whether any of them will replace VSCode.

@Hakari Do you think your evaluation of them will hinge on which would be theoretically most acceptable as a replacement, or which might have the most use as a supplement, or some other criterion?

For example, Cursor is just VSC with one sidebar you can't customize. That could score very high for someone who just wants to keep using VSC, or very low for someone who plans to be using VSC anyway.

@marvingardens I will use both of them for separate projects. There's some proprietary stuff where I don't want to transmit the entire codebase out into the world via prompts - those I will keep to traditional VScode.

The vibe code IDEs I'll likely use for brand new projects and less sensitive stuff.

Criteria...hmm:

1) Ease of use (ease of ramping up basically)

2) Does it support all the model providers I want or is it more gated?

3) Does it support locally hosted model inference without me having to bully it?

4) Do the Pro plans cost a lot? Last I checked all the Pro plans were restricted by competition to <=$20/month, so if that's still true then this point is probably moot.

5) Do any of them support generous free tiers?

6) Does it stick to my prompting or does it...deviate from parameters regularly?

7) ? Wildcard.