Posted by meander_water 1 day ago
I believe I would end up enjoying life more while everything gets solved through agentic development.
If I compare this with my previous companies (publicly valued companies), I would definitely be on vacation while still delivering perfect PRs and features in a matter of hours, instead of the 2-3 week timelines it took me and my team in the past.
Genuinely: how? Share some insights. So far I have seen anything exceeding about 2k lines of text in the context diverge. Meaning, throwing more LLM at it rapidly baloons the size at the expense of internal coherence. Things become stale, hallucinated, duplicated and outright faked. Only thing that works is constant manual intervention and pruning of the generated slop.
Not only these things are unable to find "best solution", they are unable to find any solution. Instead (especially for Opus) they seem optimized to convince user that the task is accomplished.
My investigation skill can create one, two, or dozens of probes with benchmarks. Once you're confident in the solution, you can plan and author the investigation, then execute it phase by phase.
- Post https://www.minid.net/2026/7/14/how-to-automatise-with-ai
- GitHub repo with examples of skills and rules. https://github.com/meerita/monorepo-nextjs-golang-rust-pytho...
Whatever miracle maker you're talking about, if it exists, should be making you money so, it's either doing that, is a weird train hobby, or a delusion.
I'm still surprised sometimes at how rapidly I can prototype something. Is it always shippable? No. But something that would've taken me an amount of time I can't afford to is much quicker.
This past weekend I decided to build an iOS app as a personal project. Normally I wouldn't have the time to work on one like that but I was able to get a working POC that is actually usable for my trivial use case in a matter of hours.
I'd call that a victory.
People writing software have been automating others away for a long time. The "I'll replace you with a short shell script" attitude. It's not surprising nor especially unfair we get the same treatment.
I don't see any kind of AI-assisted coding as making anything. You didn't build that iOS app, Claude did. Maybe you can delude yourself into feeling some kind of pride or contentment in it and somehow think you built it but a lot of us who actually have a passion for the domain can't.
Imagine you're a carpenter. You've spent decades honing your skills, countless incomplete or borked projects that were recycled for newer better ones, the skin on your fingers and hands have become tougher and rough like sand paper due to how the majority of your waking hours are spent manipulating wood. Now, someone gives that carpenter a machine that you can tell it your idea for any wood object and it just spits it out in a couple minutes and charges you a few bucks. The carpenter says, out of curiosity, "make me the roof in the Chapel of San Idelfonso", and after a little while, it spits out a nearly perfect copy of that magnum opus. Any real carpenter would probably be disgusted, or at least feel a deep void inside them and nothing but disinterest in the machine. They'd just go back to their workshop and leave the machine outside. But that's probably because nobody becomes a carpenter to chase the current hype thing and to make money; they have passion, just like our engineers used to.
For guys like those you can't really make an argument that they don't care about the craft when they use power tools 90% of the time normally.
So are LLMs a short cut? sure. Does that mean people using them don't care about the craft? not necessarily.
I'd love to spend more time actually learning iOS development. It's just that I work in a different area for my regular job and when I get home I don't have the time (due to various responsibilities) to sit down and spend adequate time on iOS.
The jump from writing code by hand to having a machine learning algorithm compute code for you is not the same as that. We had tags, syntax highlighting, auto-completion, word suggestion (via dictionary), intellisense, enhanced auto-completions via snippets, natural language processing, language servers, etc. etc. throughout the decades and an LLM is still not a Star Trek style replicator.
Now, it might have taken a blacksmith a hell of a long time to work various kinds of metal to order, but then someone got the idea of creating a mould so you could make more of them at once. Then later on technology advances and you can now machine it to ridiculously high precision with CNC. Now the person who writes CNC code for a high precision part might not be the same person who has to work another metal part manually.
Did the machines do that or did they? And why exactly does that distinction matter? Half the code I write depends on stuff hundreds or thousands of others have built before me.
I understand the sentiments, genuinely. I do miss the feelings of carefully crafting a great piece of code. There was a genuine sense of accomplishment there. I still get a sense of accomplishment from agentic development, but it's different. It feels more similar to when you've steered a team of people who are actually doing the work, and then the final product gets delivered. I never touched the actual work, but my input / steering / taste / direction / etc. was a big part of what was delivered. And now we have the solution that we wanted.
If you got into this field because you like to hand-write code, and that's what you view as your craft, then yeah, I think you probably are in a tough spot. There's always the opportunity to do that on your own time, just like I'm sure the vast majority of woodworkers today are doing it for fun, not as their profession.
But if you're in this field because you have a passion for building things that solve problems (you know, engineering), if that's what you view as your craft, I don't think there's any shortage of opportunities to practice your craft.
But what is code for? I don't know that I think there's any inherent value to code for its own sake. Just like I don't think a tool's value can be separated from what you do with it.
I'm not sure where the "profitable" part comes from. I know lots of people who have vibe coded little apps that solve a problem for them or their friends and family. There's no profit involved, but it solved a problem.
Can you give an example of a problem worth solving for which there is truly no demand? I don't understand that, because to me it seems like the demand is what makes a problem a problem. Otherwise it's just a state of being, and if there is no demand to change, whether yours or anyone else's, what makes it a "problem"?
But it doesn't have to be so brutally outcome-based as you're demanding, although you do get a ride on a steam train out of it too.
It's not entirely unprofitable, either, if that matters. Niche.
The AI hypers are more about “building things” than “solving problems”. When you analyze their comments and projects, they can barely articulate what the project’s purpose is about. Or even if it can be used by somebody else. It’s always about LoC, coding speed, and specs complexity.
Most of times people would say the do solve them. Still I might or might not be right that quite often they do not in reality.
Regardless, in my case, I build a SaaS app that I sell to real non-technical small businesses in a tiny niche that helps them with a particular problem they face in their business. They don't care that it uses AI, they just need this problem solved at an affordable price so they can focus on other things. This solution that I sell would have been absolutely impossible to provide in an automated way pre-LLM, and I could never have built and grown this business on my own (in this time frame anyway, probably at all) without agentic development.
So, I don't care. On the contrary: I can at least see a silver lining thanks to strong LLMs. They often can reduce the code to 20% without loosing anything obvious. Yes, it's error prone. But the untested and often not even linted code by my coworker is a maintenance hell and has a lot of subtle errors.
The main problem is that there isn't enough manpower to do proper reviews.
Which leaves us two options without LLMs: ignore the code and leave a lot of necessary work undone, or run brittle code, that never saw testing, linting and a second pair of eyes, in production and hope for the best.
At our place, I wouldn't mind if any commit would get at least a mandatory (surrogate) review by a decent LLM. Could not be worse than now.
But the damage that LLMs have done to the social aspect of work is horrible. Beyond horrible.
It shouldn't be surprising that if you swim against the tide by doing things the old way, you are going to feel more burnt out because of it.
Some people are too quick to drink whatever kool-aid they’re served. That’s all I see in these tired excuses.
> It shouldn't be surprising that if you swim against the tide by doing things the old way, you are going to feel more burnt out because of it.
Better to swim against the tide than letting yourself get swept up in the current and drowning.
It no longer surprises me, but the profound lack of empathy shown by AI (or any other popular hype du jour) proponents is worrying. A large numbers of fellow humans are saying “this thing is negatively impacting my life” and the response is “lol, and you should be thankful”.
What part of "job has changed" is kool-aid, or an excuse?
> Better to swim against the tide than letting yourself get swept up in the current and drowning.
If you really think AI is something you can "defeat" by being stubborn then best of luck to you.
> It no longer surprises me, but the profound lack of empathy shown by AI (or any other popular hype du jour) proponents is worrying. A large numbers of fellow humans are saying “this thing is negatively impacting my life” and the response is “lol, and you should be thankful”.
Not sure how this relates to my comment even a little bit.
Never get sidelined by your tools again
Here for the 3am deploy
We're watching in real time the conditions that created the Luddites.
We may never again have as many "software developers" as we did in early 2020s. That shift is sure to have severe economic, political, social, and cultural consequences.
I am especially interested in how the work in security, assurance, performance, and such hard problems will evolve in respect to the large amounts of LLM-generated code. I see a lot of opportunities there. Not in terms of how many people will be able to work on that but in terms of how much income those competent enough will command. Individuals and teams. There are some really interesting developments in spaces that have historically been shunned by the framework configuration crowd. It seems to me that the advocates of speed over quality in software development are about to be swallowed whole by the tide of code they unleashed.