I used GitHub Copilot for fourteen months. It is a genuinely good product — the inline completions are fast, the chat sidebar is useful for explaining code, and the GPT-4 integration added real reasoning capability. But after auditing my dev toolchain costs and thinking carefully about what telemetry I was feeding into Microsoft's training pipeline, I decided to stop renewing. Here is what I found and what I switched to.
The Issues with Copilot
Cost is the obvious one: $10/month for individuals, $19/month for business, with no meaningful offline option. Telemetry is the less obvious one. By default, Copilot sends code snippets to GitHub's servers for model improvement. You can opt out in settings, but the default is opt-in, and many developers never change it. If you are working on proprietary code, unreleased features, or anything under NDA, that is worth thinking about carefully.
What I Use Instead
For code completions: continue.dev with a local Ollama model (Qwen 2.5 Coder is excellent for this). It runs entirely locally, integrates with VS Code, and costs nothing per query. For larger coding tasks and debugging: Skales' Lio AI feature, which lets me describe what I need in plain language and get working code back, plus the ability to run shell commands and see output in the same interface. The quality on most daily coding tasks is indistinguishable from Copilot — and the privacy picture is completely different. Read more about Skales for developers.