On Text Editor Related Tools and Their Effects

2026-11-18

As I’ve endured LLMs taking over all facets of the internet and peoples’ minds, I’m struck by how important it is that human endeavors retain friction. Friction is what spurs us into problem solving. Friction frustrates us, but its presence leads to reward and accomplishment once overcome. Yet from the viewpoint of marketing departments, friction should be removed regardless of cost (financial or otherwise). Society has been fooled into thinking that convenience and a frictionless experience is objective improvement over prior practices. Don’t bother calling a travel agent, have ChatGPT put the itinerary together for you. It’s a waste of time to build a data-backed chart for tracking progress at work, just use Claude to create one everyday. Don’t waste your time writing documentation, Copilot can answer all your questions.

You get the idea.

Societal pressure to rely on an LLM for everyday tasks (rather than my brain) has spurred me on to reconsider the tools I depend on for my day to day work and computing needs, many of which reduce friction themselves. I don’t use LLMs, so I’m already good there. Today let’s focus on language servers, their relationship to text editors, and other text editor related tools.

Language Servers

Language servers are programs that read source code. They provide language features (such as auto-complete, “go to definition”, find all references, etc.) for one’s text editor or IDE to use. Microsoft started the Language Server Protocol so that language servers and code editors had a standardized format for communication. No longer were you limited to using Eclipse or JetBrains if you wanted intelligent language features for reasoning about your project.

These features can be extremely useful especially for those new to a project, or new to a programming language! Receiving instant information about a function/method signature as you’re typing sounds useful. Having the ability to quickly open a new window in your editor for a class definition is helpful. One can reason about a codebase quickly.

I started exploring and using language servers in early 2023 when I switched from Vim to Neovim. Exploring LSP was the main driver for my using Neovim. I was a happy Neovim and language server user once I got it all setup. I felt so swanky with all this magical new power at my finger tips. Wow, code that can understand my code?! I’ll be unstoppable.

Fast-forward to early 2026; LLMs are everywhere. Every third programming YouTuber is claiming that we never need to write code by hand again. Let your agents do it all. Watching the “AI psychosis” phenomenon spread across the software industry, my peers included, was alarming. Such a deep dependency on these magic word machines had quickly developed. You could tell that less thinking was occurring, replaced by “the chatbot said to do it.” I’m trying to prevent this post from becoming an AI rant, but I do want to provide context for why I developed a curiosity to step away from development tools of convenience. What effect would that have? Any? I paid attention to my own tool usage.

To my surprise, the language server – especially auto-complete – was affecting how I programmed. In the moment I would be thinking less about the code I was writing, instead typing a few letters then waiting with my empty brain to be presented with options of how to continue. I’d be less familiar with the layout of my own projects, because the language server made it a breeze to jump around files with no consequence. I had become dependent on the language server for my projects. That’s not to say that I could no longer write code, but I was more detached from the overall workings of the project as a result of the language server features. The LS took care of the details for me. Why did I need to care.

But I wanted to know and care. I turned off language server auto-complete. That helped but I was still more distanced from my code that I wanted. The visual pop-in of other language server features continued to maintain control of my focus and vision, rather than the logic of the code in front of me.

Treesitter

Disabling text editor features like LSP and auto-complete will poorly segue us into my next consideration: Treesitter. Treesitter is a library/ecosystem(?) of source code parsers for reading code to provide editors with better highlighting and text selection semantics based on a language’s given features and syntax. This is cool, and it made my source code have more colors, but what did it really provide me? I spent hours writing some code to install and maintain treesitter parsers for my minimal Neovim setup because nvim-treesitter was no longer maintained 1. There is also the fact that many Treesitter parses are just random git repos of code spread across the internet that Treesitter installs and compiles on your machine. The potential for a supply-chain attack is high.

The nail in my Treesitter coffin was trying to build the latest version of their command-line tool for Void Linux across all the different architectures Void supports. Rust cross-compilation is a nightmare, I now know. As of writing, that pull request is still open because I haven’t found a solution. At this point, I was ready to move away from Neovim to use an editor that doesn’t rely on Treesitter 2. Unfortunately Treesitter has become somewhat of a standard (in my opinion due to bandwagoning) so that doesn’t leave much left to chose from.

Kakoune

Enter Kakoune. Kakoune is a Vim-ish editor adhering to the Unix Philosophy. As a result, one gets a editor that is lightning quick and doesn’t reimplement any features that already exist elsewhere. Plus it doesn’t rely on Treesitter. Bonus! Additional to the non-features espoused by Kakoune, I have taken a strong stance to use the editor’s default settings as much as reasonably possible. This may change over time but as if writing this post, my Kakoune configuration is considerably small. For comparison my Neovim configuration and custom code was over 1000 lines of Lua.

Conclusion

It’s important that we repeatedly evaluate the usefulness of the tools we use, to ensure they’re solving the stated goal and to regularly guarantee that its positive effects outweigh its negative effects, but also to check how they affect us as the humans using them. Having taken my own journey over the past months from fancy LSP-laden Neovim to a stripped down Kakoune, I can say I’m happier and more engaged than I was before. I don’t move quite as fast as I did, but that was never the point 3.

I encourage you to occasionally pause to ask yourself, “why am I using this? Is is truly helpful to me? Has this affected my thought processes in any way?” Taking your own journey of evaluation and self-reflection may reveal that nothing needs to change. Or you just might discover a more rewarding world when there’s a taste of added friction.

  1. Nvim-treesitter has since been unarchived and continues to receive maintenance and code changes. ↩

  2. Sadly there is no intention for Neovim to allow disabling of Treesitter. ↩

  3. Though we are sometimes fooled, especially by YouTube tech personalities, that doing a million things per second is of utmost importance. ↩