How subtle additions of AI features to IDEs made AI features non-negotiable and necessary

When the IDE was primarily a code-editor with tools for generating boilerplate code or snippets, executing and debugging an app. And for all other features, you used a plugin.
It was a time when each line of code put into the IDE whether typed/copied from the documentation, blog, tutorial, stack overflow or other reference projects was intentionally done with full consciousness.
In that time, the only thing people weren't paying that much attention to was the inner workings of the packages they used. Need some pdf generation function, install this library but you still needed to create the integration using the library's documentation which still included some intention unless it's a direct copy and paste. Even then if people did something braindead especially if it failed, they'd take a look and try to understand what went wrong and spend hours finding a fix but if it was successful they'd treasure it like a new creation, eager to understand how it works and make it better.
There was a time when people were curious and excited about the important details and people did it according to their cognitive load.
The rise of agentic AI devalued curiosity and the need to ponder about details. The trend shifted from micro to macro, from internals to the larger implementation.
A developer walked in both worlds but at the pace that they could keep while retaining context of what they we're doing.
Now, we are encouraged to focus on the features instead and away from the specifics. Writing code has been made cheap and expendable, therefore not worth the time and effort to care for as much as before.
Did developers suddenly decide to be lazy and just offload critical thought processes they wouldn't have had before or was it a careful psychological manipulation done by the industry.
I believe that the subtle addition of AI to our IDEs without consent through updates paved the way for making it a part of daily life or work the same way that git is integrated into our workflow.
Before, people were intentionally choosing to find the AI tools that could help them generate code like Cursor, Claude, Windsurf, Lovable, v0 but suddenly an AI autocomplete then chat box then agent/planning mode was added to one of the most popular code editors in the world, VS Code.
Whether they admit it or not, the very presence of this tool on the sidebar is already a silent advertisement to test and try out this new feature and we did.
Thus the start of the agentic loop, the endless game of
1. prompt
2. test
3. revise/correct
It became addictive, there was so much that could be done and so much question that can be answered in an instant. A dopamine dispenser.
People got drawn to try it until it became "necessary" in their workflow along with marketing and expected FOMO from other people riding this train.
So was it the developers fault that they have now become dependent on AI?
Aggressive marketing campaigns from the tech industry drove this behavior forward and I also am feeling this need to delegate to AI more and more for the sake of productivity and efficiency.
So the tech industry really manufactured this phenomenon just as version control became popular through Git. Even students in colleges that weren't using Version Control Systems (VCS) before now rely on Git and Github as a remote repository of their projects.
Not necessarily. The truth is based on my opinion, agentic AI started out as an advanced autocomplete from technologies like IntelliSense, then a Chatbot that could actually understand you and then it became able to write it's generated code from the chat box.
I think of AI really as a sophisticated program with a vast knowledge-base that can parse through human language / syntax and generate the requested response which in this blog's case code.
Having a knowledge-base that's basically a collection of all our knowledge and learnings from the time of who knows when till now is a treasure and I'm glad that we have it. I'm also glad that we have agents that was already being done before via scripts, pipelines, boilerplate/templates generators, and code snippets as the pre-cursor. Agents were just the next frontier of those tools. We've been wanting to automate writing code for a long time since Assembly I think but solutions just came in different forms in those different eras.
As a matter of fact this website app was written purely with AI through Cursor using Tanstck Start https://tanstack.com/start/latest.
The main problem I'd like to point out in this blog is that it's been imposed on everyone without consent especially in the new VS Code installation where the AI features are installed and visible by default.
Thankfully, you can opt to hide or even remove them through the VS Code's official guide https://code.visualstudio.com/updates/v1_104#_hide-and-disable-github-copilot-ai-features which I found through HackerNew https://news.ycombinator.com/item?id=46501220 so it isn't that easy to find with a normal search (because it's a changelog announcement I guess)
So yeah, if it's installed and visible on launch it's definitely showing a vibe that it's part of the IDE by default part of the new developer experience. Even if you didn't plan on trying it out, you'll still be curious and that's all they need to start you on the path.
Yes, so what's the point of this? now that almost everyone's using AI for something and for developers coding.
I think it matters because we should have a say whether we opt-in to this free trial of AI tools like Copilot or just install it like a normal extension or plugin.
The problem is it's not that simple to opt-out, they don't exactly hand you a button, you have to search for it.
My main argument is that we should just be given a choice at the start of installing our previous favorite IDEs that didn't have AI before instead of being conditioned like it's now a normal part of the process. AI isn't needed to write code but it helps us therefore it should be like a normal plugin. Even Git or a VCS in general, is not installed in VS Code by default and the AI Copilot or whatever AI tool should be too.
As much as I appreciate all the new things and new possible approaches to coding solutions we can see or simulate and use in our projects via AI even automate the work we would manually do ourselves.
I would still like to preserve a dedicated space for me to code by myself with no AI assistance or agent in the very software I'm using to write and run the code. I believe despite code becoming cheap, programming in itself is still an art form, a creative expression of thoughts but I know coding is only one way we demonstrate programming because forming an algorithm just needs the mind. So even if AI does take over coding, we (people) still have the steering wheel, the idea of original thought on what we want to build.
I honestly got tired of relying on Agents recently and would just like to build at my own pace, experimenting, making, mistakes, fixing them, figuring out why without instant answers from AI though google search does not make it easy but maybe that's what we have to accept in the sense that our way of getting knowledge through search on Google has been improved from the pure website recommendations. So is that also the case for AI? is it an improvement on handwritten coding? I leave it up to you.
Regardless, spend some time writing your own code again in your free time and enjoy it like playing with Lego.
I'm thankful we still have VS Codium https://vscodium.com/ and Sublime https://www.sublimetext.com/ right now. Even Zed is offering AI subscriptions on its editor.
Share this essay

Developer, aspiring writer and content creator, music enthusiast, and a follower of Christ. A person who likes to think about many kinds of topics.