I Returned to GitHub Copilot After Switching to Cursor

I kept personal projects separate from employer and client work, so I subscribed to GitHub Copilot myself. My early memory of it was the original VS Code extension: AI-assisted completion that felt like IntelliSense taken much further. Useful. Impressive for its time. Still, in my head, Copilot stayed that autocomplete product for longer than it should have.

Then I forgot I was paying for it. The reminder arrived as an email confirming another year had been paid, and the renewal made me look again.

That period sits somewhere around my increasing use of Cursor and the move towards a more expensive Cursor plan. The exact chronology is fuzzy.

The return felt more familiar than I expected

The renewal was enough of a nudge to try Copilot again in VS Code.

By then I had been living in Cursor for a long stretch. Going back to VS Code with Copilot’s agent chat felt surprisingly close to the workflow I already knew. My reaction was not “this is still the autocomplete toy I remember.” It was closer to “this is much nearer my Cursor day than my old mental picture suggested.”

That does not mean the extension had been autocomplete-only for the entire period I was ignoring it. It means my memory was stuck on an earlier version of the product, and the return corrected that.

Having already paid also influenced the decision to use it. I am being candid about that. Starting a personal project partly because the subscription was already sitting there is not sound financial advice. It is what I did.

A self-study tool I still would not promote

The project was a separate personal SaaS idea, broadly a self-study tool. I thought it could have practical uses. I hoped that, eventually, it might cover the subscription cost. That hope was more motivational than proven.

As I understand the implementation now, it is basically fully built. It still needs testing, bug hunting and refinement before I would promote it.

I spent an intensive long weekend on it: a Thursday public holiday, then Friday off. By Monday I had reached my available Copilot usage limit.

The entire SaaS was not built in those four days. I do not have a clean inventory of what landed during that weekend versus what already existed around it. The weekend was concentrated, productive and ended against the included allowance.

What I liked, and what I missed

I enjoyed using Copilot.

The familiar VS Code environment and the agent chat made the transition easy after Cursor. I was struck by how far the tool had developed from my memories of its original form.

I cannot point to a major product failure that drove me away.

What I can say is more subjective. The available model choices did not suit me as well. The speed felt slower. I found myself missing Cursor.

The limit, the temporary Cursor plan, and then no time

When I reached the included allowance, I was prompted to enable additional paid usage. I did not want to open that billing path.

Instead I bought a separate personal Cursor subscription for one month, around $20, and later cancelled it. That temporary personal plan is not the same as the main company-funded Cursor subscription I use for ordinary development work. I mentioned both personal cancellations briefly when writing about cost reviews; this is the fuller Copilot side of that story.

I subsequently cancelled the personal Copilot subscription too. Not because I decided the product was bad. Because I became too busy for another side project. A large work project and substantial overtime ate the spare capacity. The practical decision was clearer than the invoice: I stopped trying to maintain the side project simply to justify a cost.

Looking back, with an inconsistent fallback

If I am honest, Copilot impressed me enough that I can see myself using it if Cursor disappeared.

If you read my Claude Code post, you might have caught me saying something similar about Claude. Apparently my emergency replacement for Cursor depends on which article you are reading.

I am comfortable leaving that contradiction visible. Both tools impressed me. Neither replaced Cursor in my everyday work. I still prefer the familiar Cursor setup for the day-to-day continuity I already know how to operate.

A forgotten renewal, a surprisingly familiar agent workflow, a long weekend that hit a limit, and a side project that lost out to real overtime rather than to a bad product.