Grumbling about GenAI isn’t new. Growing up as a member of Gen-Z, the constant ‘answer’ to every problem I raised or had from every parent, teacher, or mentor was always the same: “It’s the damn smartphone.”
And while it’s very easy to dismiss such a statement (the statement itself is dismissive), it unfortunately holds some truth. The advent of smartphones created widespread issues like shortened attention spans, worse sleep, and various addictions, to name a few. All this is the tradeoff, the cost, to devices that allow us to be connected like never before. Mobile devices changed the world we live in, and changed the way we speak to each other.
All that is to say, as the world adapts and adjusts to Generative AI, I’d like to point out that we have been here before. There are remarkable parallels between the moment ChatGPT opened for public use and the frenzy after the release of the first iPhone. I fully expect that by the time I have kids, my go-to phrase will be, “It’s the damn AI.”
So, since this is not our first rodeo, I want to use this perspective to analyze issues cropping up in this new world and explore what we can do better this time around. I’d like to dive into how it impacts the way we learn, skill regression, and the trust we have in our sources of truth. From there, I think it’s important to examine what benefits do come from it all that make it worth using. And finally, I’d like to discuss ways to improve personal usage of these tools, and encourage responsible and active stewardship of both individual AI user-ship and personal growth.
The Problems
Similarly to how smartphones lowered barriers to communication but raised barriers to connection, I believe generative AI has lowered barriers to entry in many fields but raised the barriers to learning. Once you discover how easy tasks can become with it, it can be hard not to immediately become overdependent on it — to “vibe code”, so to speak. By removing from you the responsibility to figure out how to do something, it also takes away both your active attention and your agency. This prevents you from really learning from your work.
This mindless flow also easily allows for skill regression. What you don’t use, you lose. If you ask for every little thing to be done for you, you will eventually start to forget the nuances and skills you do know and have. Not to mention again, by removing agency from the process, you also take a step back from the problem-solving seat, and becoming too used to not actually solving the problems will lead to a form of skill regression as well.
And last but not least, the tools are not infallible, but the speed and confidence with which they output often makes them seem trustworthy. And, as a result, when they are wrong, discovering this wrong is harder (lack of trust that self knows more than machine). This becomes a time sink, because generative AI is very often wrong.
So, given the parallels between smartphones and AI, what lessons do we already know to combat these problems?
The Prepwork: Weighing the benefits and drawbacks
Before we begin, we have to know what the problems we have with these tools are. When addressing a phone addiction — or any addiction really — it’s important to step back and look at what you both gain and lose from time spent there. What better way to do so than classic pros and cons. For phone usage, that reflection might look something like this:
Pros
- Social Media is the only place I can keep up with disconnected friendships
- Cool apps and games that are unique and fun exist there
- Modern messaging tools are more effective methods of communicating on larger scales
Cons
- I spend ~25% of my day on my phone
- Screen time causes other important tasks to get sidelined
- Too much brainrot makes my head hurt
For GenAI, we can do the same thing! Here are my pros and cons as an example.
Pros
- Explains my own logic back to me from the code written, which helps me find logical flaws
- Saves time on repetitive tasks
- Reduces the barrier to entry in new technologies/unexplored territory
- I enjoy drafting ideas with it before beginning my approach
Cons
- Too much “hey Cursor can you do ___” and I lose my ability to understand what is going on
- One misunderstanding/deviation from the genAI can lead to hours of wasted time if not handled actively
- My own learning stagnates
- The speed/confidence of output is hard for me to follow/pressures me to accept it as higher fidelity somehow
The goal of an exercise like this is to look at why you use these things in the first place, and identify where they become harmful. That flipping point is the best place to start implementing solutions.
Solution 1: Practice active participation.
Now, in the same way that doom-scrolling as a passive act is a time-sink, I’ve coined the term “doom-coding” as its blank-minded, constant prompting equivalent. Doom-coding is a level of GenAI assisted coding that takes all responsibility of the solution out of your active, problem-solving brain and puts it into the relationship between the prompts you create and how AI interprets them.
The solution to this is both simple and hard: you have to practice the skill of active participation. Take the time to really sit and read what the GenAI outputs, and find places to question it, if you choose to use it at all.
Here are the questions I’ve been asking to maintain my active focus:
- What would asking a bot about this give me that I do not already know how to do?
- Does this answer actually match my question?
- How can I improve on this?
- Is GenAI good at this type of problem?
Solution 2: Set time/access limitations.
Another classic toolset we already have includes time limits or even full-on phone access limitations. Parental controls have become extensive, and not just limited to benefiting concerned parents. For now, we don’t quite have lots of the same tools yet, but here are some possible limitations to avoid falling into the doom-coding trap.
Set yourself to “Ask-only” mode.
Most AI IDE’s have the ability to flip between an “Agent” and an “Ask” mode, where the Agent will run wild and modify whatever while the Ask remains just a simple chat window open to the right. You can still use the output, you just have to actively decide to instead of passively allow it to edit your work. This forces you to enter a dialogue instead of granting it permission to do whatever and constantly cleaning up its mess.
Set an app time limit on the apps you use for code generation.
Time limits don’t just exist on phones, there are many ways to set timers on any computer application as well. Maybe you’d benefit from being cut off from the tool entirely for a while to break the habit formed and create a healthier relationship with it. Or maybe the time limit will force you to be more purposeful in using the tool, and pressure you to optimize it for what it actually is good at instead of what you wished it could do for you.;
Solution 3: Get off the phone/AI.
As we’ve established, sometimes the best thing to actually do is step away, stop using, and to go try something tangible and real. It is always worth it to tangibly map out a complex problem yourself to make sure you understand it, whether that’s on paper, a whiteboard, or another application. There are also always people who are way better collaborators than any generated text will ever be! And maybe in some cases the Generative tools really don’t serve you at all, and it’s better to quit them until you find something that benefits more than hurts. Sometimes the best thing to do really is to just walk away.
The future has AI in it, whether we like it or not. It’s not our first rodeo with such a tool either — smartphones have paved the way as life-changing but dangerous technology. In the same way that demand for better phone health led to tools for better screen-time management, I hope awareness of how GenAI usage becomes unhealthy can lead to a better relationship with these tools in the future.