We should be using Artificial Intelligence (AI) to augment our lives. Not as a cheat code for it.
No, I’m not a Luddite.
Let me begin by stating that I am a daily user of AI.
This post is not a list of reasons not to use AI, because I think we should. There are obvious benefits to using it, and in terms of our own potential, it can have enormous benefits. Giving people more tools to learn and grow at a pace they previously believed was impossible, without extensive training or private tutoring.
There are definitely positive reasons and arguments for using AI.
However, there are negatives to its usage. There are changes in my own work that I have recognised over the past few months, and what I mistook as progress might actually be decline.
This post is about creating a healthier relationship with AI. A relationship that could greatly benefit our daily life.
Never-ending FOMO.
Social media feeds are now teeming with daily posts from designers, developers, marketers… declaring how they have maximised their workflows via rapid and extensive adoption of AI tools. Offloading tasks to multiple agents, freeing up their schedules, and improving their profits.
You’re not ‘10x-ing’ if you’re still doing things the old way. Why spend hours designing, developing, and launching applications when it can be done with a few prompts and some (soon-to-be-costly) tokens?
It’s demoralising and enough to make you want to leave the profession. Which I came very close to.
Some of these claims might be true, but others could be exaggerated. However, what is certain is that there is a push to use AI more throughout our day, without thinking about how it could affect us.
speed !== progress.
Completing a task feels great. Whenever we do that, no matter how small that task might be, we get a dopamine hit. AI usage can help speed up the delivery of those rewards by offloading our daily work tasks to a chatbot. We’re using far less energy to complete the tasks, but we gain a similar reward in the form of dopamine.
Looking at it from the other side of the coin, you can see why this approach is appealing.
Without AI, many tasks would take a significant amount of energy and thought. When these tasks are in progress, we’re using energy to complete them. Either physically or mentally, but both have an exhausting effect on the body. Depending on the complexity, you may even need to lie down after they are completed.
You’ll get your dopamine hit, but you may be too tired to appreciate it.
Is this speed a signal of progress?
Personally, I don’t think it is. It’s a shortcut.
What we are doing is actually tricking ourselves into believing that we are making progress and advancing our skills. In effect, hijacking our internal reward system.
The lure of the quick, easy fix.
During my day as a developer, I’ll have access to various AI tools that contribute to workload management. Some of these will be used for code completion, some for code reviews, and others to develop entire applications. Whatever helps to get through the increasing pressures from clients and project managers.
Take GitHub Copilot as an example. This agent runs in the background and predicts what I’m writing, offering suggestions or code blocks that, a good percentage of the time, are exactly what I’ve intended to write. This can be very useful. Particularly for developers who often write boilerplate code to get an application up and running. It gets the tedious things out of the way so you can focus on the more complex actions.
It’s often the selling point proponents of AI tell us. It frees us up, and we can spend more time on complex tasks.
But I don’t think it works like that, and I’ve noticed changes in my own abilities which have made me question how damaging AI overuse might be.
I don’t feel as confident as I did previously, and am too reliant on AI instead of taking the time to work out a problem, which I would’ve done in the past. Recalling certain functions or methods has become more of a chore, and my ability to describe what I want to do has dwindled. Even my vocabulary, which has never been the most extensive (I often just swear instead), feels weakened.
Could the overuse of AI be damaging our minds in later life?
No such thing as a free lunch.
We have AI apps installed on our phones, desktops, and even within our cars. They are now fully ingrained in our daily lives. At some point during your day, you’ll use AI either to quickly answer a question or take control of a complex task. Shifting the mental exertion from ourselves to a cluster of servers.
Researchers call this “cognitive offloading”, and it is becoming part of a growing number of academic investigations into its effects. An article written in 2026 by the BBC highlights this – AI chatbots could be making you stupider.
Some of this research is still in its infancy and awaiting peer review, but early indications show some intriguing results.
In one study, participants were asked to write a short essay. Splitting them into three groups: using ChatGPT, Google search without AI summaries, and finally, without technology.
A total of 54 students took part in the research. The students who relied on AI/chatbots showed brain activity reduced by up to 55% compared to those who had no input from AI or search engines. Their activity indicated a brain that was “on fire”.
It’s not just the brain activity that highlighted the differences. The essays produced by those in the chatbot group were “soulless” and so familiar that the students could have copied one another. This group also struggled to recall information from the essays because they had no ownership of their contents.
There is ongoing research trying to understand if this could lead to cognitive decline in later life, potentially increasing our chances of developing Alzheimer’s. Similar to the research with student essays, it’s in the very early days and awaiting peer review.
So if “cognitive offloading” is potentially harmful, what are the alternatives? How can we use AI in a way that supports us?
Aim for augmentation, not surrender.
In Black Box Thinking (Matthew Syed, 2015), the book describes AI use in healthcare and the challenges it faced when trying to get surgeons to accept the technology and its decisions. This was written before the emergence of Large Language Models (LLMs) like ChatGPT and the commonplace use of AI.
It describes how surgeons quite simply didn’t trust AI to make the same life-or-death decisions they could. The years spent educating themselves. Long, exhausting days working under severe pressure in an endlessly stressful environment. Being required to think on their feet at a time when others would crumble.
Their place in the medical profession was hard-earned, and they weren’t ready to surrender it that easily.
Syed’s suggestion was to reframe the relationship between humans and AI.
Instead of offloading tasks, AI should be working as a partner to medical teams. Never having the final say, and never attempting to control the situation. It should be offering guidance. Providing ideas that the medical team may not have considered.
AI and LLMs are outstanding at pattern recognition. They can spot patterns in a time that we cannot get remotely close to. It can research thousands of documents and sources to provide options, workarounds, and solutions in minutes.
What it doesn’t have is real-life understanding. It can’t comprehend our emotions or how teams perform under pressure. It can read research papers on the subject, but it will never have the insight that medical professionals have gained through hours of hands-on experience.
But pair these, and you have the potential for medical teams to be less error-prone and improve patient outcomes.
The aim should be for AI to augment our daily lives. When we hand over control to AI, we begin to lose our own, hard-earned abilities. But when we recognise the powerful ways AI can teach us, we open up a number of possibilities for advanced learning.
A practical example using everyday tools.
Grammarly is an app I think many of us will be familiar with. It’s given many verbally-challenged people – myself included – the ability to write confident-sounding messages at the click of a button. All we have to do is look out for the underlines, accept the changes, and in a few minutes, our jumbled mess reads more like something a professional with a decade of experience might have written.
Or does it?
We’ve already seen that research suggests AI-sourced content is strikingly similar in a small group of students. Now make that group larger. Let’s say, an entire nation. All with access to the same tools and potentially, all sounding the same.
So, how could you change this outcome?
Instead of blindly accepting the changes Grammarly suggests, take the time to understand the areas where you are consistently failing. For me, it’s often using the simple comma or hyphenation. Parts of the English language I probably should have learned long ago, but had previously just accepted my lack of understanding was part of my identity.
With this feedback, I can search for the relevant information and learn where and when the correct usage is.
Sure, this takes longer. But this is a marathon, not a sprint. And I think we have forgotten this simple statement. Learning should never stop, and AI gives us the tools to have our very own personal tutor with a curriculum tailored to our weaknesses.
A lightbulb moment.
I’ve been interviewing recently. And to be quite honest, it’s been demoralising.
These interviews have highlighted huge gaps in my knowledge. It’s felt like the world of development has shifted around me, and I’ve taken too long to notice or update my own skills.
I seriously considered leaving the sector at one point because the required level of understanding to get to a new baseline seemed too great to achieve without retraining entirely.
Fortunately, I’ve changed my mind, and I have one of the recent Durness–Inverness–Durness trips I’ve been making to thank for that.
We’ve recently sold our house, which has meant my fiancé and I have been living at her father’s on the West Coast of Scotland. An outstanding place of natural beauty and peace, especially if you want to clear your head of noise and get a moment to think. The drive up helps too, where you’ll encounter pretty much no digital radio signal once you’re past Lairg. A good time to be alone with your thoughts.
The time to think has allowed me to think about books I have read in the past. One of those came to mind during a drive back up the road. Peak: Secrets from the New Science of Expertise (Anders Ericsson, 2016). If you’ve not come across this book, I highly recommend it.
Peak delves into the science behind human achievement. What we believe is innate talent is actually the combination of surroundings, dedication, and, of course, practice. But the type of practice being undertaken is the most important thing.
You just need ten thousand hours.
You will have no doubt heard of the ten-thousand-hour rule. If you practice something for that number of hours, you will eventually become an expert in that field. Roughly translating to a decade of practice, it’s a nice round number for us to anchor to and aim for. It was popularised in Outliers (Malcolm Gladwell, 2008) but is often misunderstood.
The number of hours was an average.
Gained from participants attending the Berlin Academy of Music, Ericsson found that by the age of 20, elite violinists had completed, on average, ten thousand hours of deliberate practice. Structured learning that aims to achieve a goal in each practice session.
Complete a sheet of music with zero mistakes a set number of times, for example.
But here is the takeaway about this “rule”. Some of the participants accumulated far fewer hours than the average, while others clocked more.
It was also domain-specific. These are elite musicians, and what might apply to them may not apply in other areas or sectors.
However, what does apply is deliberate practice. Ericsson refers to it as “the gold standard”, and it’s what we should all be aiming for when learning a new subject or expanding on others.
Developing a personal curriculum.
With the (recalled) knowledge of purposeful and deliberate practice, I’ve started to wonder how AI can augment my own learning.
Development is often a solitary existence. Many hours are spent learning new methods, performing tasks, and upskilling with languages and libraries. I’ve been a developer for over a decade and suspect I’ve reached that ten-thousand-hour mark. Possibly surpassing it.
But we know that number is not as important as some claim it to be, and if I look back, not all of those hours were structured. This is where my approach to learning needs to change.
My personal choice of chatbot is ChatGPT. Within the app, you can separate chats into projects, and in each project, you can include sources for the AI to reference. Here, there’s a spreadsheet that lists what I feel my current developer skills are. Using a 1-5 grading system – grade 1 being the lowest – I outline the areas I’m comfortable with and those I believe I’m weaker in. I’ve also graded the languages or topics on career priority, using a similar scale.
Prompting the AI, I first check that it understands the concepts of purposeful and deliberate practice as outlined by Anders Ericsson, which it does. Then, I instruct it to act as my personal tutor and outline the reason for learning. What I hope to achieve, how I would like to be graded (1-5), and how feedback should be delivered (highlighting errors with descriptive corrections).
With the relevant sources and prompting information, ChatGPT has been able to create a syllabus tailored to my current knowledge and provide me with tasks that sit just outside of my comfort zone. Perfect for purposeful and deliberate practice.
Instead of creating lessons that start at the very beginning of a topic, it takes all the areas I want to learn and works out to be proficient in a certain language, I may only need to be comfortable with a few core concepts. With those basic concepts, you can often transfer that knowledge to other languages and libraries, as long as you understand the primary syntax.
It really just starts from there. It’s a dynamic curriculum, and I’ve already modified it as I start seeing how to get the most out of it.
How it works in practice.
I’ve been using this setup for the past week, and so far, I’ve been very impressed.
Whilst it’s far too early to tell how much I’ve improved, what I can say is that the structure of learning and the “stickiness” of subjects have improved due to the repetition required for each task.
A little more about that.
The learning is set up as daily tasks, focusing on a particular subject. As an example, the first day involved confidently describing what happens when using an async function in JavaScript and the resulting Promises. As well as writing an error-free async function to retrieve posts via the Fetch API.
It’s the error-free part that is important.
One essential part of deliberate practice is having a goal. A target to aim for which allows you to confidently say “I’ve completed this task”. In my case, it’s repeating the process from start to finish until the code is error-free and matches the challenge.
I’ve mentioned that I use GitHub Copilot in my development environment. For this type of practice to work correctly, I’ve turned off Copilot completely. Everything is about recall, so I can’t have any predictions which could give me hints. This repetition, from start to finish, even when there is a minor error, appears to help create a much better understanding of the task and its structure.
Each session takes around two hours, so enough time for something to be challenging and not so long that the tasks become tiring.
Once I’ve completed the session, I’m graded, a review of the task is outlined, and details of the following day’s task are provided. I’ve also set ChatGPT to update my developer skills document so it can continue to cross-reference my current knowledge with what it plans to teach me.
But the main takeaway here is that AI is teaching me. It’s not doing the work for me. It’s exactly the setup I’ve needed to prevent “cognitive offloading”.
The final word.
Overall, I’m happy that I’ve allowed my mind some time to think. It’s helped me get over a hurdle in my career and understand that, like any tool, it’s how you use it.
AI had given me the feeling I was slowly being pushed out of a job I enjoy. Many of us, myself included, have used AI to shortcut our lives, and in later life, we might regret that decision. By choosing AI to augment our lives, I think we’ll have a more rewarding relationship with the technology.
It’s going to be fascinating to see how far this could take my learning.
