The Development of AI
2022 — ChatGPT sparked the generative AI wave
On November 30, 2022, OpenAI launched ChatGPT, allowing users to have conversations, ask questions, and generate content using natural language. This brought AI closer to the general public.
2023 — AI advanced in reasoning and multimodal capabilities
In March, OpenAI introduced GPT-4, which supported text and image inputs. In December, Google launched Gemini 1.0, further accelerating the development of multimodal AI.
2024 — AI could see, hear, and speak
GPT-4o was introduced with near-real-time voice and visual interaction capabilities. OpenAI also introduced o1, focusing on advanced reasoning, while Google developed Gemini 2.0 with AI agents in mind.
Source: GPT-4o · OpenAI o1 · Gemini 2.0
2025 — AI moved from answering questions to taking action
Models such as Gemini 2.5, OpenAI o3, and GPT-5 continued improving their reasoning, coding, and tool-use capabilities. AI agents gradually became a notable trend in workflow automation.
Source: Gemini 2.5 · OpenAI o3 · GPT-5
2026 — AI became an important tool in the workplace
AI is increasingly integrated into programming, research, data analysis, and business automation. According to the Stanford AI Index 2026, AI capabilities continue to improve, but reliability and safety remain challenges that need to be addressed.
Source: Stanford AI Index 2026
In just about four years, AI has evolved from text-generating chatbots into systems capable of reasoning, using tools, and helping people carry out complex tasks.
How AI Has Affected Me
I was actually pretty slow to get into AI.
When I first joined my company, all I knew was that AI could help me research things faster than traditional search methods. However, its answers could still be wrong, the code still had bugs, and I still had to check and fix things myself.
After about two months at my current company, I was assigned to a research project focused on applying AI to our work.
At first, I didn't understand anything about these workflows. Hooks, skills, and all sorts of other concepts... I had no idea what they were. It wasn't until much later that I finally understood what AI agents were and started realizing what they could do.
From that point on, I gradually fell into the habit of relying on AI without even noticing it.
From reading tasks and fixing bugs to writing documentation, almost everything in my workflow involved AI. Instead of trying to understand a problem first, I started asking AI for help as soon as I ran into difficulties.
It wasn't until I watched a video by Pham Huy Hoang (@toiicodedaoblog) that I started looking back at how I had been working.
I realized that my ability to think things through, my habit of exploring problems on my own, and some of my programming skills were being used less and less.
It's not that those skills had completely disappeared. I was simply getting used to letting AI handle things that I should have been thinking through myself.
I spent quite a lot of time thinking about this and about what I needed to change.
I don't think I should completely remove AI from my workflow. Instead, I need to learn how to use it properly, so that AI supports me rather than becoming something I depend on.
I'll probably talk more about this in another post.
How AI Is Giving Some Developers "Brain Rot"
There are two people at my company whose ways of using AI sometimes really annoy me. One is my boss, and the other is a senior developer on my team.
Why? Because, from my perspective, both of them seem to trust and praise AI a little too much.
My boss uses AI for almost everything. He's pretty good at prompting too, with things like: "Imagine you're this kind of expert or that kind of expert..." before asking AI to solve a problem.
I have no problem with using AI to brainstorm or explore ideas.
But sometimes, I've seen him ask AI about professional matters even when he's sitting right next to the people responsible for those areas.
I remember hearing him prompt AI right in front of our Marketing team lead.
At that moment, I couldn't help thinking: Wait, why not just discuss it directly with the person in charge? Maybe a conversation with someone who has real-world experience would offer a better perspective?
Sometimes, my boss also shares news in the company group about company A laying off employees to adopt AI, or AI solving problems that took humans years to figure out.
I understand that he might be trying to encourage everyone to stay up to date with technology.
But if we only look at AI's achievements while ignoring the context, limitations, and the role of humans, I think it's easy to develop an unbalanced view of what AI can actually do.
The second person is a senior developer on my team.
He's also very interested in AI and uses it frequently at work.
What bothers me isn't that he uses AI, but how he responds to other people's opinions.
Sometimes, he brings up a problem for everyone to discuss. People on the team offer suggestions, mostly based on their own experiences and situations they've dealt with before.
But instead of analyzing those ideas together to see which ones make sense, he copies their responses into AI and asks it to argue against them.
I don't think there's anything wrong with using AI to explore different perspectives.
The problem is that if AI's answers are always treated as the basis for rejecting everyone else's opinions, then what's the point of having a discussion?
He wants to improve and level up, but from what I've observed, he sometimes struggles to accept feedback from the very people he works with.
Over time, I've started feeling a little out of place at work.
When phrases like "Just brainstorm with AI" or "Make the most of AI's power" keep coming up, I start wondering: Are we using AI to work more efficiently, or are we gradually handing over the thinking itself?
Are we paying less attention to the skills that are essential for developers, such as debugging, reading and understanding code, analyzing problems, critical thinking, and designing solutions?
I don't think AI automatically makes those skills disappear.
But if we ask AI about every problem, hand every new task to AI, tell AI to fix every bug, and merge code as soon as it works without really understanding what's going on, then what are we actually learning from that process?
Another thing I often notice in tech communities is the FOMO whenever a new AI model, framework, or skill comes out.
Every time there's a new technology, people want to try building this app or that website.
There's nothing wrong with that, of course. Experimenting is also a way to learn.
But sometimes, after all the excitement about how powerful the new technology is, the final product turns out to be just another todo list, note-taking app, or website with features similar to hundreds of existing products.
I'm not denying the value of small projects, especially those built for learning.
What concerns me is that we're becoming too focused on how quickly AI can create a product, while paying less attention to what problem that product actually solves.
In the past, when we wanted to build something, we had to think about the problem, the design, the implementation, and the trade-offs involved.
But now, sometimes the first question is: "Which AI tool can build this the fastest?"
2026 — AI agents drive a surge in programming activity
According to GitHub (October 6, 2026), developers and AI agents made 7.38 billion commits in September 2026 alone, more than five times the number recorded in September 2025. Overall Git activity also more than doubled over roughly a year.
Source: GitHub Engineering (2026)
Looking at these numbers, it's clear that software development activity is growing rapidly, with automated systems and AI agents playing an increasingly noticeable role.
But I don't think the number of commits fully reflects software quality or a developer's actual productivity.
A project with thousands of commits isn't necessarily better than one with only a few hundred.
Similarly, when browsing repositories with lots of stars on GitHub, I've noticed that many AI-related projects are attracting a lot of attention. Meanwhile, some repositories that I personally consider useful and well-built receive much less attention.
Of course, the number of stars isn't an absolute measure of a project's quality either.
But it makes me wonder: Are we becoming too focused on what's new and what AI can generate, while spending less time thinking about the actual value of a product?
These days, there are more and more discussions about AI replacing developers, the job market becoming more difficult, and which skills will still matter in the future.
I don't know how much AI will change the programming industry, and I don't think we can conclude that AI will completely replace developers.
But there's one thing that worries me.
If we become too dependent on AI without continuing to develop our own thinking and technical skills, we might be making it harder for ourselves to adapt to future changes.
Final Thoughts
These are just my personal opinions and experiences from working, using AI, and following different tech communities.
I don't think AI is bad, and I don't think vibe coding is wrong either.
Especially for people who aren't programmers, AI can help them build apps or websites for their own needs and make their work more efficient.
I also know many highly skilled people who use AI effectively and appropriately. I've learned a lot from them as well.
What I'm trying to say isn't that we should or shouldn't use AI, but how we choose to use it.
If we trust AI too much and believe that simply combining ourselves with AI means we can solve absolutely anything, maybe we should reconsider that mindset.
Not because AI isn't capable, but because figuring things out, making mistakes, and solving problems ourselves are also part of how we grow as developers.
AI might help us finish our work faster today.
But if one day AI was no longer available, would we still be able to understand, evaluate, and solve the problems we're working on?
That's also a question I've been asking myself.
P/S: This is my first blog post, so I'm sure my writing and perspective still have plenty of room for improvement. If you have a different opinion or a similar experience, feel free to share your thoughts! ^^