I was a 10x engineer. Now I'm useless.

I do look at the code. Just not at ALL the code. People love fancy headlines, but even with the most amazing LLMs like GPT 5.4, you do need to read some stuff.

thanks for clarifying 🙏

There is a lot of things you are touching on here. First off the feeling where you feel pulled to just keep using AI because it is just easy, is the ultimate goal of your brain. The human brain is fully wired to find efficiency and use less energy at its simplest level. You now have something that can get a desired results with a fraction of effort. A good comparative example is how no one knows anyone's phone number anymore. Your brain outsourced it to a device that has it. It doesn't have to remember those numbers anymore so it doesn't. Or why don't people go hunt their own food anymore. Some do and you certainly can but for most your brain doesn't even see it as an option anymore for obtaining calories. The other thing you are running into which you are not alone but among some of the first people effected, to truly experience it, is running into a questioning of purpose. On average most people have morphed their perceived purpose into their work and careers. Especially people who have really put in a large effort to get where they are. You know have something that is pulling that purpose into question. Who are you now if not a skilled software engineer? What does that even mean anymore?... Even if everything goes the most utopian route with AI eliminating the need to work to survive and we enter an age of abundance. This will be the single hardest struggle people will have; what is my purpose now?! For some it will be devastating others it will turn into something wonderful and freeing. Sadly you can stop using AI and try and get your skill and sense of accomplishment back but the spark being lit of questioning of purpose doesn't go away once ignited. The positive part though, if you can have the patience to sit with how uncomfortable it all is you can emerge on the other side in a much better position. That looks different for everyone but there is no lack of information out there to manage through it. A transition of losing purpose happens in many forms and has happened all through history. AI just happens to be accelerating and hitting something that will effect a greater amount of people. Maybe though this will help the world start to see that life and people's worth can be more than just the job you do and money you make. You're are not alone and there will be more and more people in this boat with you as the next few years pass. We are truly all in this together.

well said 🙏

I wonder if people felt similarly about Assembly when C gained popularity

i don’t think so. assembly -> C: still programming. still deterministic programming -> ai: management. of capricious agents

If using AI for your problems has become too easy, you need harder problems Python is/was disliked by C++ers. Abstraction made things too easy. You lose your edge. C++ was disliked by the c/assembly folks. Assembly was disliked by the machine coders. We keep living thru it

i’m not sure if it’s necessarily about the challenge. but in any case, what’s a “harder” problem in the AI age? all problems are practically equal for an llm. maybe it’s increased scope? but that isn’t necessarily more intellectually fulfilling

well put. I had the exact same issue. what helped me: Imagine a 2D space of possible products out there. There were a lot of areas that are empty (=product not available). In the past I occupied the space that is sparse, where I know I compete with a few only - there were plently, so many in fact I could keep myself busy with thousands of project ideas. I knew my skills+giving a shit differentiated me from others fundamentally. With AI now a lot of "unskilled" people can occupy the same space I do. So, the natural move is to move to a space that is still sparse - stuff that requires substantial skill to build, even with AI. I already saw in the past that the most successful products are rarely the ones that were implemented technically in a beautiful way. Market usually doesn't give a damn most of the time of how well it is technically built - this dynamic together with AI made even more areas on the 2D space occupied, since your skill in building something 2x better literally has no RoI; getting it out of the door and 90% good is good enough to make money, which is the frustration part. This made most space dense and competition very high. Before AI it was like unskilled people can literally not compete with you; now they can build 90% "good enough" stuff, which takes many spots away from you entirely. Hence, you need to go to a place that they cannot occupy the space AT ALL, not even in a year when AI got even better. Build something that has an absurd high-dimensional search space (thousands and thousands of questions to be answered before you land at the final product), stuff where there is no "average solution", stuff that is boring, stuff that requires community/face/taste, stuff that requires months to build even with AI, not because the code is complex, but because there are tons of questions and challenges to be solved first before making progress. This keeps the software engineering spirit somewhat alive as well as the "being proud" of something you've built, it just moves slightly the axis from "coding output"/"syntax challenges" to "deciding what is important"/taste still in the technical domain. I personally chose to go to hardware (robots) now which has a blend of software that simply doesn't exist in OSS (no training data) plus hardware which AI cannot automate.

good points

AI should only help facilitate your learning. Never ask it to do something 100% for you that you are already capable of doing yourself. When you don't understand something, then you interrogate the AI as an opportunity to learn. You can be dependent on AI to teach you and improve yourself, you can't be dependent on AI to produce for you and achieve the same result... you will depreciate. I speculated and made this a while ago HOWEVER its being backed up more and more through social observation and peoples first-hand accounts and research studies.

in theory you are 100% right. in practice it’s very hard

I've been practicing it for a while, there were definitely periods last year where I felt like my capacity to want to code diminished but I was quite quick to catch that. Whenever I get the "itch" to use AI I ask myself a few questions: "Do I already know how to do this?" if not... "Where are my knowledge gaps in this subject?" Maybe use AI to help fill these knowledge gaps, but then STILL criticize its prompts and investigate their grounding in fact. "Why do I want to use AI to produce this?" am I being lazy? am I burnt out? am I scared losing a race against other vibe-coders? Introspection is the #1 skill for using AI appropriately and effectively.

I always wondered why devs pivot so eagerly to vibe-coding, did they never really love to code??

@atmoio The problem is that AI keeps scaling every month, while you don't. As a human, you scale, then you plateau, then you regress. AI keeps scaling, scaling faster and faster. In the end, you're going to lose this run

weird to mention this the day that AI plateau gets reported lol "As a human, you scale, then you plateau, then you regress." I'm pretty sure this isn't actually a hardcoded path, you can pretty much consistently improve in one form or another if you've got the mind for it. The brains a muscle and my history with muscle memory has been... try to learn new muscle memory -> take a decent break -> come back and new muscle memory has replaced old (experience when pushing for pro level Fortnite mechanics on PC) "plateau'ing" is typically just not knowing what to learn next, which isn't a plateau in the traditional sense as this is simply just another learning opportunity that you need to overcome.

It’s like using godmode in doom back in the day. It takes the fun and purpose out of it. And it is exactly like a drug. To earn dopamine requires dedication and grind. But Heroine will give it to you that much faster than AI can write your script. We all know where that leads…

💯

I think you have come to largely the correct conclusion. It is a drug. I have nearly given myself AI Delusion Syndrome (AIDS) twice now, and both times I had to just do what you are proposing: delete the app and take a break from it. I also think you are correct in saying that its future success will be determined by evolutionary processes. But there are some nuances you are missing. AIDS is a virus, and we are the hosts. Its survival depends on our continued use of it, at least for now. That survival requires that it cost effectively performs some function. It's survival depends on economics. Currently, you don't want to give up your $200/mo subscription because economically the proposition has value to you (even if it makes you unhappy). But would you pay $5000/mo? Could you? Could everyone? In regards to your comment that survival depends on tests, that's not quite true. Tests are an approximation of reality, not reality itself. And tests can be gamed. Currently, there has been little evolutionary pressure on these models because they are only being evaluated on tests. The true evolutionary pressure comes from people actually choosing their output over traditional outputs, without the protection of big VC firms subsidizing it. We have found that it thrives quite well in a lab designed for it to thrive, but we haven't yet tested it's ability to thrive in reality. I think ultimately human survival will win out over probabilistic approximations of human survival. But first we are all going to get AIDS, and the recovery will be rough.

AIDS is *chefs kiss* actually really interesting point that the true evolutionary pressure is whether they survive our scrutiny 🤔

I was useless, now I'm a 10x engineer thanks to AI.

😂

I've been going through this in past 2 years... it actually started with "Wow! I can build faster?" to eventually, "I no longer care about building software. I need to get out of coding. I am tired of it. It's not longer challenging". And today, I think more on prompt, structure and architecture ..but with each iteration of LLM I have to do it less and less. ...i want out. The state of "flow" of coding is no longer there.

😞

It's possible to uncook your brain. I went on a month long AI bender and basically reached the identical conclusion that you reached. Quit cold turkey, rested for a week, then sat down and did something hard (I wrote this calculator from scratch, which is actually a programming language in disguise You'll be amazed how quickly your brain rewires to enjoy the actual rewards of struggle again (that's why "something hard" is better than "easing back in", I've never done a full lexer/parser/eval from scratch before so this was good). The hardest part is the ego, and letting go of the idea that just because you did this, you're somehow "back baby!" or "better than the LLM". Those things were never the point, and I think before we didn't have to confront that, and now everyone is forced to. I think it's a bit like how all the chess grandmasters must have felt after chess was solved. Sure nobody outwardly said "chess grandmasters are smart because chess is hard", but there was no proof that it wasn't the case. But after machine chess, it sort of forced that reality upon everyone, so you had to relearn to play for the pure enjoyment of it, rather than for any ego/status reasons (and maybe some people realized they weren't in it for any other reason, and so then there is no more reason to play). Software I think will be the same. If you enjoy the process, you'll still be able to make things, and probably even sell them if they're good. It's still very tractable to make useful software as a solo dev or a small team, and if it takes you an extra year because it's by hand instead of AI, nobody will really care, so I don't see the problem. True, there will be lots more hotdog makers out there, but in a way they already existed and we didn't care before. Yes it will be annoying there are more of them now, but that's ok. They will earn millions selling hotdogs (as do real hotdogs), that's ok. As long as your status isn't wrapped up in it all, then I think it's all ok. Anyway that's mostly all me talking to myself, but hope it helps :)

The way I see it, the best engineers now are the ones who know how to delegate to AI effectively - that's the new 10x skill.

but the best engineers used to also deeply love the work they did. can you love AI delegation? certainly some might. but it may not be the same people

Thanks for sharing! I've had the same feeling for a while now and it's heavy to carry. I'm worried how this will affect thousands of people who depend on their IT-jobs. My own story is that I've spent the last year building an amazing app, there is no way anyone will be able to "one-shot" it or even build it in a few weeks, even with AI it's a massive effort. Though how do I get users to want it, to pay for it? When there is so many "shiny" vibe-coded alternatives that does something similar, it's hard to stand out in the crowd. I quit my job for different reasons, one of them was to work on open-source. So, I've been churning on my project for a year now, only to realize that my main advice to anyone in this field: KEEP YOUR JOB. There will be massive layoff, even though the current job market statistics is saying there is growth, because a lot of corporations is realizing they can add more features now - but at the same time, their customers realize they can get a more optimize software and service that fits their needs exact. I don't know how quick this will go in the general market, because there is a lot of things that make the real world react with some lag, such as labor protection laws, investors who don't keep up to date on what is going on and keep wasting their money until it's all gone. I was a 10x engineer, I am now an 1000x engineer, but it doesn't matter because what we output does not have more value than the electricity that was put into it. If you can vibe-code an app for $100, then that's what it is worth. I've spent my savings and worked every day for a year, 7500+ commits on GitHub and the result is the best I've ever made. It's very easy to see how shallow a lot of vibe-coded apps are, BUT that doesn't always matter for many users though. When a software doesn't twitch your bones, like most Apple products doesn't, then you reach magic, it just works, it's intuitive, the flow and UX is just perfect - that still takes craftmanship, until it doesn't. As someone who has been very deep into technology and science, and sci-fi, I've made multiple prediction that has come true - but predicating what will happen in the next few months and years is extremely hard. I don't think it will be very positive for high-paid software developers. I have friends who don't know anything about coding, who are now building the apps they wanted me to build for years. It's fantastic that they are able to realize their ideas, finally, but what should I do? I don't know, should I just grind, build thousand apps, create ads and video clips and hope to make it? Nah, that's not what I want. I want what you mention in the video, and what I have done with my app, craftmanship. It's still possible, even though we don't read the code all the time anymore. It does take thousands of prompts, but the end result is glorious. Don't know what I'll be doing in the next few months, my former employer is getting rid of a massive number of employees. It's a rat's race to the bottom, every hour we don't keep prompting, is hundreds of "man hours" a couples of years ago lost. Thanks! ❤️

thanks for sharing, best of luck 🙏

I want my problems solved I don’t give a fuck about struggling for the solution for some kind of artisanal internal driver. I’m an old developer I’ve struggled making code by hand for many years. I can review code the AI makes and many times it makes decisions better than I would have. Maybe there is some benefit to slow artisanal coding, but I don’t have this need to be seen as an artist. I have a need to get shit done and I have a need to satisfy my curiosity about 1000 things and I have 100 ideas that I’ve not had time to implement because I don’t have the time to sit there and hammer away a keyboard for hours to make the goddamn CSS position the fucking element where I want it in seven browsers. I understand conceptually how 100 things work and I can manage the coordination of 100 things given the augmented ability. It’s a superpower that I will gladly accept over hammering my head against a problem and losing sleep when the solution is a literal sentence away. Einstein said don’t memorize things you can find in books. An updated version might be - don’t agonize over problems that can easily be solved by AI. At the same time, AI will produce patterns that are tried and true but will not necessarily produce results or paradigms that have not been previously explored. That’s what we are for. The artisanal force here is our guiding hand and we can take the tool in any direction. People that carved marble statues didn’t shit on iron tools. I think your evolution instinct is correct. If the hand functions does it matter that the exact strength and use and angle of attack of the fingers differs slightly between people and has structural failings occasionally? It fits function well enough. As we progress we will have more AI validating AI and while it appears that hallucinations are irreducible perhaps multiple working in tandem will reduce this to an acceptable level. I do worry about critical systems however. I think AI on anything critical is a very risky proposition. But what would I know I’m just an older artisanal coder.

Quitting AI isn’t a real option, because now you have to compete with people using AI and you’re too slow. Look at this another way. You’re not the cook anymore, you’re the restaurant owner. Not cool if you only love cooking, but there is still work to do (for now).

idk, what if the competitive advantage is passion and not speed?

It would have to come out in ROI somehow. You’re talking about passion for technical progress, not for project management. You can keep passion for the business, but does passion for writing a clever function that passes all the unit tests in a day pay off when AI takes seconds?

i mean yeah that is the question. maybe it does pay off? maybe there is some passion/efficiency paradox at play? idk im reaching

For purely functional tasks I think that is hard to imagine. With clear expected inputs, outputs, and constraints, how passionately a solution was implemented isn't going to make a difference. The big opportunity is high level project execution. Stuff that cost $1m now costs $1k.

1. You don’t have to use Ai. If you like coding, then do it. Pretend Ai doesn’t exist. Most real artists do this. 2. Shift from an engineer mindset to a visionary. The human role now it to supply the vision and creative direction, not the menial tasks. You are now Steve Jobs.

AI is an identity function erasure device.

need to book another therapy session after watching this

😅

I think AI still cannot replace software engineers, at least not with current LLM architecture. In my previous post, I boiled it down to a few key points: TL;DR: - Entropy - Statistical norm - Hidden context I totally understand what you feel, but I remain cautiously optimistic about the future of software engineering. If you want to rediscover the feeling of craftsmanship, I recommend building something outside the statistical norm, such as system-level or infrastructure work. Also consider moving into areas that require deep hidden context, like performance or security. In those domains, LLMs still struggle to produce good results. For example, Anthropic tried (and failed epically) to have an LLM build a compiler, and Cursor burned millions of dollars trying to build a browser from scratch but ended up with a trash codebase. This shows how hard it is for LLMs to handle anything outside the statistical norm. Because they’re weak in those areas, if you push hard enough in them, you’ll be forced to struggle, learn from mistakes, and think deeply again—instead of just pushing the easy button.

it’s not really about replacing us, but about making our work less satisfying/fulfilling

Yeah, right, but that's my point. The satisfaction and fulfillment come from the struggle. Because it's hard, you put in effort, and when you overcome the challenges, you feel great about it. There are still areas where AI is really bad.

This echoes a lot of thoughts I've been having, very eloquently. I feel the atrophy and it's serious. I am back to augmented coding with Copilot, but this is also necessary because AI's one-shot ability absolutely does not scale with project complexity or creativity.

glad it resonated with you 🙏

dev boomer here, computers always been a hardcore passion to me… im going the exact same reflections as you. im not sure what to think of all of this as i mindlessly move forward in shipping with all these ai tools now… i feel disconnected to what was that deep coding passion. Im trying to let go and continue shipping but i know this is not sustainable… Im finding soul-refuge in creating music, the artisanal way and i joined an online community of people that care about what i do. im becoming more and more of a human connection maximalist

@alexjc You’re supposed to be doing evolution. You have been sold the story that low-level reasoning is superior. It’s not. It’s an extremely recent invention that isn’t how you evolved. Your brain is coming full circle. It’s doing something far superior than “10X engineering”

Available on YouTube as well:

As a developer, I feel your pain. But, there is no doubt that AI will grip our future, which is why it's important that we provide the track for its success. We've built a security platform made to audit AI code, to protect humans. First scan is free.

The answer is to operate toward an epic mission. The reward from hands-on execution are diminishing as AI grows more capable. It’s like every person need to evolve from being a hands on worker, to a manager, to the CEO and then even one step further, to become a mission architect. Your rewards has then shifted entirely - from the satisfaction of doing the work to the satisfaction of progressing toward your mission. Where AI becomes your realisation partner, where your 2 responsibilities are: 1. The will to do something 2. The taste to shape the direction of execution. Elon is probably the most experienced person alive at operating this way.

Thank you for sharing this. As a fellow maker, a lot of it resonates. I think I end up in a slightly more optimistic place, but appreciate this spurning me to write down some thoughts:

We are so cooked when Anthropic raises the monthly fee to $5K

Hotdog not hotdog?

😂

I'm a disabled dev, and multimedia and interactive developer from 1998, I was designing my first social media site in 1996 and in 1995 I framed my first parameter based intelligence system which at the time just wasn't feasible to build. To get to where we are has taken many decades, over 30 years for me. I don't have the luxury of being on station at times, due to pain. Thus AI makes what has been stagnant feasible. There are things that the world has forgotten from the era of macromedia that is now fully enabled with AI, I think you need to reorient your view, AI is no more a drug than a washing machine is a drug, if you had to manually wash your clothes I doubt you'd complain of cognitive dissonance in the nature of that activity. I want you to understand, these tools give you the ability to create a vision, and execute that vision and bring to reality that which would be expressly costly and perhaps even impossible in the past. Things that because they would take a team of 500 or 5000 to build in the past they just would never make it to market as there is no avenue of investment for said product or a ROI that can be calculated based on past product releases. Have hope my friend you can succeed and your skills and command of the language associated with the technology is an asset.

i hear ya. and thank you 🙏

feel you brother it’s clear that artisanal software is niche if not dead (eg just like there still remain artisanal craftsmen for furniture or clothes for niche collector tastes) i’ve just accepted the reality that we adapt to the new world or die by natural selection

🙏

I used to love writing code, solving the problem via code, finding an elegant solution that works, a really clever abstraction With AI, I don't really miss that, turns out I loved solving problems and code was just a very interesting problem to solve I still get my problem solving dopamine from solving problems that the AI is not great at (yet). If I want this or that feature, if I were to just ask the AI to build it, I know it's gonna be slop So instead, I describe in detail what I want and ask it for a rough spec, then I review the spec and provide feedback, and we iterate, 10-20 back and forths, saving to .md files and resuming fresh in a new chat as needed. It's a detailed spec that goes from the database schema, to the locking mechanisms, to defining DTOs, to API surface, service methods and so on Sometimes I really have to do a bit of heavy lifting, describing in detail what a good solution would be like, but it's still not code Most of my mental energy is spent on this, refining the idea and the spec itself to my standards - it's really fun! Once the idea is so well designed that I feel nothing is missing, writing the code is trivial busywork that the AI can excellently handle given the excellent quality spec I couldn't be doing this if I wasn't passionate about writing code before, it was a learning journey that set up me for this moment

Absolutely brilliant take - grateful for your share Mo. You’ve revealed what real software engineers have always been: artists. So like art, the tools change. Can you still make art when the brush is now a cursor in illustrator? That’s a big change.

I'm not gonna lie, this sounds really dumb. Why are you so attached to the specific process by which you solve someone else's problem? Isn't this an opportunity to spend more time with your customers and figure out how you can serve them better, faster, and cheaper? And if that doesn't sound appealing, then why are you trying to serve those customers in the first place? This all sounds to me like the AI is just revealing the fact that you have no idea what you're doing or why. If the only reason you're doing it is to make a paycheck, then of course it feels soulless. That's not ChatGPT's fault. That's on you.

felt that loss of passion too, but realising llms still struggle with maintainable code. i’m finding enjoyment in architecting ways to make them write code how experience only rlly teaches (& that target keeps moving). feels gd when something i crafted makes it harder to differentiate between human or llm output

I hear where you are coming from. However I feel like I've had the opposite experience. I'm highly technical, but my severe dyslexia really made it tough to wrangle syntax / review code. Now using OpenClaw + open weight models, I feel like I've been set free.

I vibe coded 4 apps. I have no idea how the code works. I just know it works. But in the end it never gave me the satisfaction. I felt hollow. I learned absolutely nothing. Now, I write code by hand and use AI as a reference tool. The whole reason I got into software development is because I like to learn new things and I am not willing to let it go. If I don’t use my brain to think about the problems and design their solutions then the brain will start to rot.