Github Copilot

Written by Brian

Github copilot is an autocompletion tool powered by the Codex language model by OpenAI. It saves developer’s time by suggesting lines or entire functions based on context from your code. It is especially useful for tedious tasks like plotting or otherwise visualizing data.

Here is the hype video: https://www.youtube.com/watch?v=IqXNhakuwVc&t=47s&ab_channel=GitHub

Here is a video on how to use it: https://www.youtube.com/watch?v=Fi3AJZZregI&t=766s&ab_channel=VisualStudioCode

In my eyes, it is especially useful for:

  1. Automating the boring stuff, like copying and pasting blocks of code and changing a few values to plot different data.
  2. Doing things that you know you should be able to do, but can’t remember how to do right now. An example of this for me is adding a vertical line to a plot. Instead of looking up how to add a vertical line at point x, I just comment “# Add a vertical line at x” and it autocompletes the next line. Easy peasy.

After using it for ~6 months I can say that it has probably improved my coding efficiency by ~ 30% once I learned how to use it well, a skill jokingly called “proompting”.

All students can (and should) get copilot for free as a part of the student developer pack. https://education.github.com/pack

For everyone else, the quoted price is $19/seat/month. Based on this price. If a person with a $30/hour salary who works 35 hours/week who spends 50% of their time coding, then the cost of that code is ~$2100/month so to justify spending $19 USD or 25 CAD then copilot needs to make them at least 1% more efficient. I think this is true in almost all cases, so I think it would be a worthy investment.