I was going to write this blog and post it on my musings page, but I thought, why not write it here instead. AI, or LLMs, have been accelerating rapidly, Opus 4.6, GPT-5.3 Codex, Kimi K2.5, and so on. With this pace, I feel coding will soon be largely automated. Caveat: nothing is ever fully automated; you still need someone to maintain a bird’s-eye view of the process. But yes, the era of writing code entirely by hand will soon be over. I think 60–70% of code will always be co-authored by an LLM.
What does this mean for the software engineering profession? I think it means fewer jobs. Most software work happens in the service sector, and with automation, a lot of it will vanish. This won’t happen overnight, but it will happen. The CEOs of LLM firms have been saying that software engineering will be automated, though they do have a biased view. I’ve used almost all the major models in the arena, and code is getting commoditised fast. I don’t have a doomsday mindset, but I am a realist. Digital services and tasks were always on the verge of being automated, and when AGI arrives, they will likely be the first to go. Digital tasks have very little leverage, physical tasks, on the other hand, have more leverage against automation.
When I was in college, I never thought about changing professions or safeguarding my job from automation. I didn’t have that mindset back then. I knew that in the SWE profession one has to continuously learn and evolve, but I never imagined I’d have to plan for changing professions mid-life. This idea of “getting automated” didn’t come from nowhere, since around 2020, things have been moving in this direction. Anyone who used GPT-2 or GPT-3 could have easily predicted that, within a few years, SWE would move toward automation. I’m still not entirely sure about the verification side of things, but I’m fairly certain that coding by hand is going away. Coding was never the difficult part anyway—it was time-consuming, sure, but not the hard part. The real challenge was problem-solving: coming up with solutions while considering trade-offs. That will continue to exist, and problem-solving skills can be used anywhere. Still, this shift is going to affect the economy and software jobs, especially in the median segment.
I’ve never thought much about my tech career or climbing the ladder. I haven’t stayed in one place long enough to think about it seriously (not that I left all places willingly). But lately, I’ve started thinking about designations and the responsibilities that come with them, though I’m still not sure how they fit into my life. I’ve never really cared about titles or promotions; my only aim has been to write good code, solve problems, and generate shareholder value. I still think there’s time left before LLMs fully replace us. What comes next? I don’t know. I’ve always gone with the flow, if things change, I change.
I have no advice; I just have opinions.
I also don’t understand the AI Agents hype at all.
I don’t think most people actually have tasks they want to automate at that level. I’m not delegating my financial decisions to a machine, and even the ultra-rich won’t let machines automatically schedule everything without accountability, which currently doesn’t exist. The idea of my AI agent talking to your AI agent for matchmaking is one of the dumbest things I’ve heard. A lot of tech bros seem disconnected from reality. Sure, having multiple agents working on a codebase sounds cool, but who is verifying all of this? If my friends find out they’ve been messaging an AI, they’ll probably disown me. My AI agent ordering food for me? I don’t know, I like ordering things myself. If all UI interaction is removed, I’ll just be bored. Food delivery and ride-hailing apps have already spent millions reducing friction, click a button and things show up at your door. How much easier do we really want it to be? If AI-agent ordering were truly valuable, these companies would’ve built it already—they have all our data. The fact that they haven’t tells you something. A lot of this feels hype-y, like geeks routing a lamp command through China and finding it amazing. Vanity is fine, but our world model shouldn’t be built on it.
Also, the whole “Moltbok world takeover” thing was kinda fake (https://x.com/gkcs_/status/2020215784434462930).
I’m just ranting, please don’t send the police to my house for criticising agentic AI. I’m trying to understand things from an economic and reality-based POV. I think AI is a transformational technology and will change the world for the better. I don’t even store coding scripts anymore; I just ask AI to generate them again. Searching takes more time than regeneration. Small models are getting intelligent fast, and we should be grateful to live in this era of acceleration, there’s a lot of good that can come out of it. That said, when I think about how AI has changed my life in impactful ways, I can’t point to much (just that my job can be in jeopardy :P). It hasn’t cured diseases yet. We were promised more, but I think we’ll get there, just not today or tomorrow. Maybe the day after tomorrow. Remember that exponential curve?
As a software engineer, you might be feeling dread. So am I. But if we aren’t adaptable, then what are we? The world is changing, and we should change with it. There has never been a better time than today. This is the printing-press moment, but for code. I think all software engineers will either become AI Engineers or Applied AI Engineers, we will evolve. The rest of the infrastructural tech will be the same. I don’t see DB Engineers getting replaced soon :’) We do overrate change, things work slowly. I remember in an interview a candidate told me that they were using an AI Agent to run workflows of exporting data. I asked why an AI Agent to do that? You can just use an Event based trigger, right? They said that it was a client requirement lmao. Anyway, if you are feeling lost then go read a book or revamp your personal website. Don’t be sad. Be happy. Buy a humidifier (send recommendations pls)
Educate. Empower. Annihilate.
