Why I Quit Copilot | Prime Reacts

The speaker discusses their recent decision to stop using the AI coding assistant Co-Pilot. They recount a humorous incident with their computer before delving into their reasons for the switch. The primary reason was a simple oversight—they forgot to sign back into Co-Pilot after setting up a new computer and reconfiguring their Neovim environment. Despite this, they have no intention of returning to Co-Pilot. The speaker also poses a question to the audience about their usage of Co-Pilot, highlighting the financial implications for Microsoft and suggesting a potential investment opportunity. They express concern about students having free access to Co-Pilot, arguing it removes the initial struggle that is crucial for learning to code effectively.

The speaker explores the adage 'if you don't use it, you'll lose it,' particularly in the context of coding skills. They discuss their personal practice of periodically turning off language support to sharpen their skills and maintain a strong connection with the programming language. The speaker introduces the term 'Co-Pilot pause,' describing how they would wait for Co-Pilot to suggest code instead of writing it themselves. They reflect on whether relying on Co-Pilot has fundamentally changed their coding approach, comparing it to copy-pasting from Stack Overflow. They also mention regaining the joy of coding without the assistance of Co-Pilot.

The speaker debates the impact of Co-Pilot on their creativity and coding style. They acknowledge the 'Co-Pilot pause' but dispute the notion that their creativity has diminished. They advocate for a balanced approach, using Co-Pilot as a tool rather than allowing it to dictate the coding process. The speaker also shares their increased proficiency with Vim and the benefits of Co-Pilot in speeding up certain tasks. They reference a GitHub study that found Co-Pilot improved developer satisfaction, but challenge the idea that it benefits only weaker programmers, asserting that the frustration of coding is often due to encountering bugs rather than the act of writing code itself.

The speaker discusses the importance of enjoyment in achieving mastery in any field, including programming. They argue that the challenges faced by junior engineers are a necessary part of growth, and question whether Co-Pilot could hinder this process by bypassing these early struggles. They also touch on the idea that to truly excel, one must enjoy the process, and relate this to their own experience with side projects and hobbies. The speaker uses the example of playing the guitar to illustrate how passion and enjoyment can lead to expertise.

The speaker addresses concerns about the quality of code generated by Co-Pilot, suggesting that leading the AI with clear instructions can produce better results. They discuss the importance of using TypeScript return types to improve Co-Pilot's accuracy. The speaker also mentions the limitations of Co-Pilot, particularly with respect to outdated suggestions due to rapid changes in software development. They share anecdotes about their experiences with Co-Pilot and the need to refer to documentation, despite using the AI tool.

The speaker raises privacy concerns associated with using Co-Pilot, which uploads code snippets to a remote server. They express discomfort with this aspect of the tool, especially considering their values around self-hosting, open source, and privacy. The speaker also discusses the limitations of AI in understanding the context of code versions and the potential for generating outdated code. They mention their intention to explore open-source AI models that can be self-hosted, reflecting on the importance of privacy and the potential downsides of contributing to the training of AI tools.

The speaker ponders the privacy implications of using AI in coding, questioning how much of a concern it should be. They debate whether the snippets sent to Co-Pilot servers could be pieced together to form a meaningful whole, and whether this would require intentional effort. The speaker also contemplates the broader issue of telemetry and data collection, and the potential for users to inadvertently train AI systems, possibly at the expense of their own relevance. They express a desire for a privacy-focused version of Co-Pilot, even if it meant paying more for a less effective tool.

The speaker reflects on their decision to take a break from AI, citing fatigue and underwhelming experiences amidst the hype around AI technology. They express a preference for a simpler era of coding and suggest a sense of disillusionment with the current state of AI-assisted development. The speaker also touches on the idea that sometimes, the past can indeed be better, and that it's a common mistake to believe the present is always the best. They conclude with a humorous comparison between different versions of a game, implying that personal preference plays a significant role in what one considers to be the best approach or tool.