Headlines like this do more harm than good. I sometimes wonder if those are paid marketing or people actually write them on their own.

He stated that "software development IS NOT CODING", but coding is part of it. Not all, but part. If you outsource that knowledge you will be forever trapped in the walled garden that is (and will get worse) those LLM services that code for you. If you are senior developer you will know when and how to use tools to help you, but people starting in the industry will get the wrong idea.

All that wave of 'X is over because a new LLM is out' is discouraging for anybody new searching for that field and irresponsible. The cynic in me tends to believe this is on purpose to build a vacuum on such markets and fill it with mediocre work.

This will take a while for most people to internalize, but now with GPT-6-Astra we are basically in the age where as a developer you control agents to write code for you. This will be further cemented by the next generation of models.

I fear that the rest of society is not at all keeping up with the pace of development, and we're quite soon going to face a situation where most white collar work will be done by agents. What this might do for employment is worrying.

> Model capabilities are evolving so fast that what was true last month may not be true today

I dunno. I think the harnesses are getting a lot better but older models still do the bread and butter just as good.

> Recently the head of Claude Code, Boris Cherny, said to delete all your AGENTS.md Claude.md files and just see what the models can do nowadays

It was pretty obvious early on (which is only like a year ago) that you shouldn’t have much outside of things the model/person has no way of knowing inside your agents.md. Basically just a pointer to your docs. That OP didn’t know this as well as they listen to number one shovel salesman Boris makes them seem unserious.

> My process for learning a new domain or technology used to be modeling it in code: build to learn

This is still the process. My guess is OP is actually just an average / below average developer, so AI causes this reaction. I also have this reaction anytime I use AI in area I’m a novice in. You still need to learn domains deeply to unlock the ability to use AI effectively. Domain dependent, but if you’re not seeing AI make mistakes consistently you need to learn more about how the tech you’re using works. You’re missing things.

The world is still full of ambiguity and tradeoffs that have no objective correct answer. I don’t think AI or humans have solved all these cases yet.

Biggest benefit of all this getting the control of hardware back to consumers. Software has been the soul of the computers which only knowledgeable enough could put into hardware for a long time.

Now (almost) anyone who owns a programmable device also controls it the way they want, as if they actually own it.

I am stretching a bit here but we are almost there now already.

I wish these models start writing super optimal code that can squeeze very last bit of the chips so that we can get rid of Electron and similar bloatware.

Yeah, the industry will quickly shed all of the people who needed it to be one way and will grow with all the people who are capable of it being a different way.

So the wheel of the market turns.

Let me make a guess, OP's job is internal dashboards

I actually think everyone should still code a little. Maybe 30-60 minutes per day. While the majority of code is generated by agents, I've found that with even a little coding I catch stuff that I miss by just doing code reviews of agentic-written code. But yeah, the majority of my day isn't coding now.

I see all these comments here dunking on the article, but I have to say, I basically completely agree with all the points the author is making. I think there are different kinds of software engineers, and the author's point of view resonates with those of us that are mostly concerned with building tools, rather than products. At the end of the day, I need the tool to work so that it can be used to build other things. In that sense, coding is dead for me. Anything more than trivial few dozen line changes to the codebase is going to be done by claude. It will figure out the hidden dependencies and gotchas more carefully than I will on a given time budget. And for me, unit tests have always been an afterthought, so the fact that the agent will also add unit tests for its modifications means it's strictly more disciplined than I ever was. I've been at this for 25+ years, and had the fortune to be able to largely be a "cowboy coder" but those days are long gone.

I still love writing code myself, but as I use LLMs more, I'm realizing it's not really the coding I actually loved, it was the software architecture and up front design work. Once the structure and plans are laid out, writing the code is just mechanical, and I'm going to reach for an agent if I can. Of course, these days, I find myself asking claude for advice on architecture as well, but that is much less a deferral of responsibilities than asking for an audit or survey.

It's hard to take the title seriously when the first paragraph ends with "It’s hard to know right now if the investment will pay dividends or not.".

I too think that probably the mechanical action of putting fingers against a keyboard to inspect, understand, debug and build software might effectively end up like writing punch cards is today. But it still irks me if we'd try to shoehorn in getting there instead of actually looking at this taking all the sides into account, including if it's actually worth it given what it costs today or will cost tomorrow.

Before commenting, it's worth reading through the entire thing (big surprise), but especially this:

> Software development IS NOT CODING

The author is talking about a very specific part of software development, not all of it, as some might understand by "coding".

Entry level coding has changed in that it can be replaced by AI. But this is a preference, not a rule or law. However, AI is still many years or even decades away from writing good complicated well thought out software. This is software that big F500 companies typically spend a lot of money on. This software allows F500 employees to outsource their work to AI since the hard complicated stuff is taken care of. One way to think about this is AI writing. Everyone agrees AI prose is shallow, repetitive, lacks creative structure, and can be slop like. By no means is writing dead. In fact, it’s the opposite, good writers have now become more valuable. This is something the author mentions.

If this is true, why don't I see a flood of high quality AI-written software? All I'm seeing is a bunch of vibe-coded shit that's basically of no use to anyone.

>but It no longer makes economic sense to write all the code by hand, so increasingly I don’t.

and

> There are good developers, mediocre developers, and bad developers, and the good ones are still hard to find.

move in opposite directions. Friction is the source of learning, which creates good developers.

undefined

> I will miss it

When honesty goes out the window...

I call BS on that, it's been 6 months since Opus 4.6 and we still do not see a revolutionary software release or a major increase in quality. At least we see some security bugs being finally fixed.

Stop peddling this nonsense please.

Coding will never be "over" - those saying it aren't true coders in the first place.

>I’m a software engineer at JPMorgan Chase. I have more than 15 years of experience in building software and I used to love it

Big woop :) I have 50 years as a professional programmer and all AI will do is create more programming jobs. Many times I have seen the latest and greatest thing that will eliminate programming. You know what all of those did ? Increase demand for programmers. The exact opposite of what management wanted. AI, same will happen once the shine wears off.

undefined

Worked on my first borderline-computer-sciency not 100% trivial task in a long while today: Not sending one SNS message per event, but batching them on the client side, sending out as batched message if it went beyond a certain size or beyond a certain timeout (-> limit event delay caused by waiting for batch to fill up.) (this will save A LOT of money)

I wrote a nice spec for it, gave it to claude fable and it completely shit the bed. I ended up writing most of it myself and let claude iron out the bugs and kinks. Ultimately it was just 150 lines and far easier to follow and much more sensible than the fully automated claude output. I don't understand why AI couldn't build this.

Apparently I am still not living in the dark-software-factory parallel world many other people live in. Am I already part of the permanent AI underclass? Lapped and too far behind to catch up?