It’s hard to escape the ubiquity of AI tools when you’re a person who uses the internet. It is doubly hard to escape when you’re a young person glued to your phone for most of the day. In a surprise to no one, people aged 14-29 spend the most hours a day looking at a screen (by a large margin, I might add). Recently, TD reported that 90% of Zoomers are using AI tools in 2026, a sharp increase from the 76% reported in Deloitte's 2025 survey. Our own Stack research on learning and AI corroborated this data, finding that 67% of early career developers use AI tools daily in their work. That’s 13% more than in 2025, and 10% more than the cross-generational average.
With so much time spent staring at lighted pixels on their handheld electric boxes, the fact that Gen Z’s AI usage is skyrocketing probably doesn’t make you gasp in shock. But the numbers don’t paint the most flattering picture of my generation, especially as those numbers continue to increase so starkly year over year. Although Gen Z is proficient at knowledge discovery via the wealth of instantaneous search tools at our disposal, what Gen Z is not so great at is knowledge retention. Finding an answer is one thing, but remembering it over the long term is another. Neuroscientists are finding Gen Z is cognitively disadvantaged compared to older generations.
And while I’m the first to argue that not all AI usage is bad, many of the effects of prolonged AI usage aren’t great. My fellow Stack writer Ryan Donovan recently reported on the growing body of research about the negative effects of letting your AI agent think for you—what smart people are calling “cognitive offloading.” This cognitive offloading distorts our perceptions of reality, cedes agency and decision-making to bots, and outsources ethical and moral judgements to training data. Basically, it’s a bad idea to let AI do your thinking for you. Combined with my generation's declining attention span and lowered ability to retain knowledge, AI tools for learning create a compounding negative effect on people my age that we’ll probably be digging ourselves out of for the rest of our lives.
This all becomes increasingly ominous when you remember AI’s non-determinism often leads to incorrect or misleading information. It’s no secret that trust in AI has declined, even as adoption has increased. It seems the more we use these tools, the more sure we are that they aren’t always right. Gen Z feels this wariness as much as anyone, yet we still use these tools daily. Part of this is that AI tools are easy—and I will only speak for myself, but I’m Gen Z and sometimes I get really lazy—but the other part is that my generation feels more compelled to use AI than any other generation, if only to get an edge in a fierce and rapidly disappearing entry-level job market.
We’ve created a culture ripe for cognitive offloading, holding our young people to increasingly high standards during their schooling before punting them straight into a job market without room for them. For Zoomers, using AI tools may be the only way to get an edge in a society that demands high performance without always rewarding it. We have to be bigger, better, faster, stronger if we want to be considered for work, offering provable value from day one that outweighs what an AI could do in our place. So even if we don’t trust the chatbots, we sort of have to trust the chatbots.
We can hardly forecast positive outcomes for the future of technology—and the world, honestly—when our youngest workers are in active brain atrophy, and getting all their knowledge from chatbots they don’t even trust. So what do we do? It’s been said a million times, but if you have no junior developers now, you’ll someday have no senior developers. Fostering the talent of the future means writing documentation, curating resources, and offering work experience. But these are all things we readily and progressively outsource to AI. How will the next generation of tech workers learn if the only ones willing to teach them are chatbots?
And what can Gen Z do to create the neural pathways necessary for long-term wisdom and knowledge while still getting the positive benefits out of AI that can feel like a prerequisite for competitiveness in the shrinking early-career job market?
I think the answer to both of these questions is the same: a knowledge base.
Stop Gen Z staring at TikTok and start Gen Z staring at your knowledge base
What is a knowledge base? In simple terms, it’s a central repository of knowledge. You can think of it as a sort of living library of all the information you know and might need someday, kept in one place for easy reference and long-term access. Stack Overflow is an example of a public knowledge base for everything you might want to know about software development, and our enterprise product Stack Internal is an example of a private, internal knowledge base that houses just the particular knowledge of a company.
But not every knowledge base needs to be collaborative or about a central, cohesive topic. My personal knowledge base has things like Stack’s brand guidelines, notes from my film projects, sewing patterns I want to try someday, and even a record of updates made to my apartment lease. What’s important about my own knowledge base is that I’ve written down, catalogued, and organized all of the information I’ve collected over the years into one place that I can easily search and reference.
The most important thing I do with my knowledge base, and what I believe will have lasting positive effects on my cognition and learning, is write notes about the things I learn. For a long time, we’ve known that humans of all ages are bad at remembering information. They even have a model for it in psychology, called the Forgetting Curve, that shows the steep drop-off in information retention humans have when that information is not actively reinforced. Basically, we lose 50% of new information we learn in an hour, and the rest of it in a week. This is bad news for anyone trying to learn anything ever.
Luckily, there are plenty of ways to reinforce learning so you don’t forget about it. Notetaking is usually the first step, followed by reviewing, quizzing, and summarizing. Often, you have to repeat these steps over and over in order for things to really stick, which is why you probably had to take so many exams in school. When you’re out of school, though, it’s easy to lose these kinds of habits in favor of the easiest thing, which is just to look stuff up when you need to know it.
But looking stuff up isn’t actually learning. You probably don’t need to burn into your memory how many gold medals the women’s US hockey team has won, so it’s OK to just look that kind of stuff up via a search bar or a chatbot. But what about GDPR data protections? Or best practices for REST API design? Or debugging common errors in C++? Or must-have security measures in your code?
If you’re asking for a step-by-step guide from a chatbot every time you need this information, or just sending an AI agent to do it for you, you’re cognitively offloading. An hour later, you’ll have forgotten 50% of what the bot taught you. A week later, you won’t remember any of it.
While the days of skimming textbooks and digging through pages of search results are gone, thus making the search and discovery part of active learning increasingly obsolete, active learning as a whole is still very possible with modern technology. This is where your knowledge base comes in.
For people my age, we have to make the active effort to reinforce and practice the knowledge given to us by AI tools. The anchor point for this is notetaking. And I mean active notetaking, by the way: no copy-paste of large pieces of text, or rewriting them verbatim from what you're reading. To keep your knowledge base manageable and actually learn a thing or two from what you’re consuming, you have to sort through the noise and record what you actually care about in your own words.
That’s part of the beauty of a personal knowledge base—if you want it to actually be useful, you have to curate and organize. You have to actively engage with it, otherwise the mess of information you’re leaving for your future self makes the knowledge base useless. Not everything you read from a Stack Overflow Q&A is going to be useful to you in the long term, so you have to make the effort to sort through the noise and write down the things that might actually be. The same can be said of answers from GenAI—so much of it is useless, sycophantic pitter-patter that clogs up your knowledge base when included. Learn to cut the crap and keep only the stuff that matters in your knowledge base.
Then use what you wrote down—not the answer from the AI—to complete your task. Afterwards, write down a few notes about what you did and where you hit obstacles in the task. Doing this will reinforce and actualize the things you were taught by the AI, helping you retain the information past the dreaded first week of the forgetting curve. Next time you need those answers, return to your knowledge base; afterwards, write a little about what was actually useful from your original notes and what new things you learned. Keep returning to your notes and updating them until you’ve mastered the learning. Every time you do this, you’re reinforcing the learning you got from the AI chatbot, making it a thing you actually know how to do and not something you just cognitively offload to AI. Now we’re really cooking.
But this still doesn’t answer what we should do about AI’s non-determinism feeding young people misleading or totally incorrect information. We simply cannot ask young people to rely solely on AI to learn, even if they are learning actively. That’s why knowledge bases aren’t just important for junior developers or college students. They’re important for senior teammates, mentors, and leadership to upkeep knowledge bases to service the development of the next generation of tech workers.
Teach us what you know, we beg of you
I grew up learning many skills from YouTube videos and WikiHow articles. It’s why I can expertly Photoshop a picture of me with One Direction, among other actually useful abilities. This is a trend amongst my age group—our 2025 Developer Survey found that 55.2% of 18-24 year olds use videos when learning to code and 60% use other online resources like forums and communities.
But this kind of content is starting to disappear from the internet. This is, in part, due to decrease in demand and overall traffic—Chegg sued Google in 2025 over their steep traffic decline, citing AI as the main culprit (RIP Chegg, you’re the only reason I passed Lit 3). Not even the major upholder of internet information and documentation was safe: Wikipedia experienced an 8% decrease in visits during 2025 because of AI. Bleeding users often means bleeding money, especially for educational sites that rely on subscriptions, ads, or—godforbid—donations. Of course, less money means less resources means less content. Less content means less people visiting these sites. It’s a vicious cycle that feeds the AI slop beast and pushes us towards an inevitable content collapse.
Asking questions and searching for answers are both major tenants of actual learning, as are repetitiveness and duplication. That’s why it’s so scary that our world wide web has felt increasingly empty, because that means there’s less human-created confusion and human-created content to dispel that confusion available on the internet. These sorts of human-to-human digital interactions are important resources for young people to learn with, and they keep our internet healthy and alive.
Human-to-human interactions in general are essential for young people to learn. Honestly, in the right environments, it’s our preferred way to learn. As discussed in our recent survey on AI and learning, studies show that students find learning from a teacher more effective than online courses. On the job, young people believe that hands-on projects and in-person mentorship are the strongest single accelerators for career-readiness, according to a KPMG survey of their own interns.
Yes, being able to provide these things to young people might seem like unnecessary work for more senior team members and leadership. But I’d argue that it is deeply necessary, lest we reach a future where there are no senior developers. Investing time and energy on young workers is investing time and energy into the future of your company and industry. But let me get off my soapbox and talk about how it’s really not as much work as you think.
When we think about all of the internet resources that are slowly disappearing because of AI—the Cheggs and Wikipedias of the world—what we’re looking at is the degradation of our knowledge bases. Eventually, that will mean the degradation of our AI tools, as they are trained on our knowledge bases. That’s why it’s pivotal that companies work to preserve and catalogue their existing knowledge.
Much of the knowledge and context you use in your daily work really only exists in your own head. Documentation and notetaking have gone out of style for all age groups, and that means the important information that keeps the systems and processes of your organization going could be gone in an instant. I recently learned that this is called—rather morbidly—the bus factor, referring to the number of people who can get hit by a bus before a project completely stalls due to lack of access or knowledge. You don’t want your bus factor to be low, just in case a bus driver in your city goes on a murderous rampage.
We, as a society, want to keep our bus factor high. We want young people to know how to do things so that people can retire or take sabbaticals or deal with sickness without our infrastructure degrading or our technology breaking. In fact, we want our technology to get better in the future, which will only happen if people continue working on it. Simply put, if we want to keep the comforts of our modern tech stack, young people need to know how to do things. They need to know how to do the things you know how to do.
That’s why knowledge bases are so critical not just for long-term organizational health, but also for the future health of industries at large. Documentation and written guides are necessary resources for young people learning the skills necessary to keep our systems running smoothly. And answering junior colleagues’ questions gives younger people important context and perspective on the things they’re learning. Luckily, both of these things can be done on your own time with very little context-switching for you…as long as you’re keeping up a knowledge base.
The truth is, knowledge bases aren’t just for young people. Skill atrophy from AI usage affects everyone, not just my generation. If you are cognitively offloading to an AI tool by having it solve problems or complete tasks for you, you are also opening yourself up to potential negative effects over time, like the erosion of critical thinking skills. But you can mitigate these poor outcomes by documenting what you know and regularly updating your notes after AI-assisted tasks. You can retain your important knowledge by actively engaging and reviewing it, even when an AI agent is helping you complete your daily work.
By building this knowledge base for yourself now, you’re creating resources for new team members and junior developers to learn from. The thing you did for yourself can help someone else. The same goes for answering questions from more junior employees—teaching concepts to someone is a great way to reinforce what you already know so you don’t lose it to cognitive offloading.
We won’t be using less AI anytime soon. You’re probably using it more and more each day without realizing it, a sort of cognitive-offloading lifestyle creep that can be dangerous for your brain if you’re not careful. Building a knowledge base can help you with that. And it can help teach a future generation necessary skills, so you can someday retire on a beach and know the things you’ve worked on will live on.
Oh, and your AI will get better, too
There’s a fun little virtuous cycle that you can create when you really lean into knowledge base building. When you allow AI to access your knowledge base, it’s able to generate answers that include your particular context and knowledge. Your knowledge base can teach your AI tools all the nuances and specificities of how you work, and you can use those AI tools to help you solve the particular and unique problems only you have. Then you can document what worked and what didn’t in your knowledge base, which feeds and shapes how your AI tools support you.
This also works well when you’re utilizing knowledge bases for the development of the next generation of tech workers. It answers the question of what we should do about AI’s misleading and incorrect information in regards to young people’s learning. You can equip your AI with the knowledge you have so its answers are generated from information you trust. Then, you can let junior team members use AI tools freely with the reassurance that what they’ll absorb from them is the same as what you’d teach them yourself.
All this knowledge base building and AI tooling can be done securely and privately, too. You don’t have to share all that sweet, sweet knowledge with the world. There are plenty of private knowledge bases that connect to AI tools through MCP, and lots of them that you can try for free. Whether you’re someone my age trying to keep your brain from rotting too much, or someone who’s trying to build knowledge resources for yourself and your AI, start building your knowledge base today. Your brain will thank you for it.