This is weird to me because I don't think I talk to anyone regularly who even uses Cursor anymore, let alone Copilot. It's Claude and Codex now, and then people with more interesting/oddball TUI agents or async web agents.
I use Copilot pretty much exclusively because it's our "approved AI solution" at the office and they block access to Claude. Since I'm so used to Copilot at work, I just end up using it all the time.
Do you regularly talk to anyone in a Microsoft shop (like most enterprises)? If you work in a ms ”partner” company there’s rarely a choice between the Microsoft product and something else. Azure over AWS is a given etc.
Many people working the front-end space seem to really like Cursor. It seems to have a dedicated audience. Myself, I have never liked VS Code so I don't have a pull towards it.
I think the costing there is the problem and with (GitHub) Copilot. Not owning their own model and not able to take advantage of the (probably brutally subsidized) fixed monthly packages that have relatively generous limits means Cursor and Copilot can't compete cost wise on a per-token basis.
JetBrains has similar problems with its "Junie" product.
Maybe some of these will have second chances once Anthropic and OpenAI run out of $$ runway and are forced to charge something closer to actual cost. Or if the Chinese open weight models do some more catching up.
Really? I use Github Copilot. It is great actually, and I actually prefer it to Claude (loosely held opinion tho).
I use Github Copilot because it's what my job provides to me. But 95% of my usage is via OpenCode (which is officially supported [0]), not copilot-cli or their IDE plugins. The rest is autocomplete in the IDE.
I actually find it to be a great deal, especially because they charge by request rather than token. So if you provide detailed prompts a lot of work can get done for very little cost.
[0] https://github.blog/changelog/2026-01-16-github-copilot-now-...