The Battle for Developer Mindshare Among AI Code Assistants

I enjoy coding, but I don’t do it like I used to. I produce more shippable code, faster and better than ever before, and find new ways to approach problems at a vastly accelerated rate. I embraced AI peer coding earlier this year and it’s been absolutely remarkable. I use it to suggest packages, write and explain code, produce documentation, solve errors, translate from one syntax to another, and even learn a new language that I might have been tempted to avoid beforehand. Of course, I’m not alone. I’m just one of what’s been estimated to be as high as over 95% of coders using AI for code assistance in some form by the end of 2023. That has big players vying for market share. Case in point, JetBrains just released its AI Assistant for peer coding in its popular IDEs, continuing a Battle for Developer Mindshare that will rage on into the new year.

For the vast majority of us, AI coding assistance has come as direct ChatGPT engagement, followed by indirect GitHub Copilot adoption. Where Copilot relied on earlier OpenAI models, it now uses GPT-4 as well. In other words, OpenAI’s models have dominated. In contrast, JetBrains uses a myriad of models, which now apparently includes GPT-4 to some extent. How these models are used and when seems to be determined by JetBrains, and not its users. Copilot has extensions for use in IDEs like VSCode and Visual Studio, but is not exclusive to the Microsoft sphere and is available for JetBrains IDEs as well. That integrated experience has been greatly enhanced in recent weeks. Copilot now includes new features that are likely to make using ChatGPT directly less attractive than installing and using Copilot’s IDE extension.

There’s a cost regardless. You must purchase a monthly subscription for Copilot at GitHub to use in any IDE. Copilot currently costs $10/mo for an individual GitHub user, and $19/mo per seat license for organizations. A $39/mo enterprise seat license is expected early next year, with certain governance features cited as justification for the price disparity. JetBrains’ newly released assistant is $8.33/mo for users and $16.67 for organizations respectively. So is it worth the price tag? Microsoft and JetBrains might privately argue that you’re getting a bargain - not just in the increase of productivity, but that your use of code assistance is largely provided on their dime right now. The compute and memory requirements for providing this kind of AI assistance has high costs and it's widely believed that big stakeholders are playing loss leaders to capture market share - especially among heavy enterprise users that hit their models hard. That argument aside, I personally can justify another $10 a month or less for direct IDE integration given the productivity lift I’d estimate to be at perhaps 150+%. Will enterprises justify price increases for seat licensing? That we’ll have to see.

Who will win the battle for developer mindshare? It’s unlikely that Copilot will lose traction immediately given its present adoption and tight coupling with GitHub where most of us manage our source regardless of IDE choice, syntax, etc. For some JetBrains users, the new assistant may represent an additional cost on top of what they already have available to them from GitHub depending on their configuration. But beyond price positioning and muscle flexing, models and recipes are advancing to provide better experiences. Learning curves that might have previously dissuaded users from switching to something new, are themselves becoming less acute thanks to the models themselves. As for me, I’m enthusiastic about the competition. I expect to be delighted.