AI that finishes your code while you type, like a prompter next to the programmer. Made by GitHub, owned by Microsoft, with the models behind it built by OpenAI and Microsoft.
In practice it looks like this: you write a comment "function that sorts customers by order date" and before you've written a single line of logic, the whole code is already suggested below, grayed out, like a shadow of text. You press Tab and accept it, you press Escape and strike it out. It does this for a single line, for a whole function, even for a file. There's also a chat mode, you ask it "why isn't this working" and it explains the error in plain language.
It exists because writing code is full of repetition. Every programmer has written the same thing a thousand times, just with different variable names. Copilot is trained from models built by OpenAI (the company behind ChatGPT) and Microsoft, on code from millions of public projects on GitHub. That's why it recognizes the pattern of your thought from the first three words of the comment.
It's used by programmers at all kinds of companies, from interns to seniors, for a monthly subscription per person. It comes in as an extension in Visual Studio Code (Microsoft's free program for writing code) and a few similar environments. For companies the point is simple: fewer hours on routine code, more on the hard decisions.
There's also a price for that convenience. Sometimes it suggests code that looks correct but hides a bug or a security hole, because it imitates style, it doesn't understand logic. It has happened to return a piece of code almost word for word from someone else's licensed project, which raises the question of whose line that really is. That's why a good programmer doesn't accept it blindly, and reads what it hands over before pressing Tab.
What's Good Here
The most accurate comparison for me is the prompter in a theater. He stays hidden, whispers the line on time, but it's the actor who says it on stage, and he's the one responsible if it sounds fake. Copilot whispers code, but the signature under the committed change is yours.
That's why I see it as a tool for speed, not for quality. It speeds up the hands of someone who already knows what they're building. In the hands of a beginner who can't judge when the prompter is wrong, it's a recipe for code that works at first glance and breaks later.
Update, 16 August 2026: the model picker is no longer just OpenAI and Microsoft. On 14 August SpaceXAI announced Grok 4.6 is available in GitHub Copilot - for VS Code, the Copilot CLI and cloud agents, via the model picker; some business accounts have to enable it first from settings. It shows the direction: Copilot is now a shop window where every big maker wants a line, because that is where millions of people actually write code.