Is coding going to be obsolete?

A specification detailed enough to compile a runnable program is called code. Even if it is in the form of a diagram. Obligatory not-xkcd.

No.

Besides, what you describe is the UML- and diagram-based programming that was popular around 2000, which failed. Something that appeared with different names before and since. In all that time, workflow engines tried to control the business processes, with various degrees of success.

Because diagrams aren't enough to describe an application, while text can easily specify those diagrams. Think about PlantUML or Mermaid diagrams - all are text files. The actual business requirements are too imprecise to specify as a single neat diagram or state machine.

Diagrams are impossible to version or diff too, making it all but impossible to control changes. Sure, all workflow engines have some kind of versioning but actually showing the differences is far harder than it sounds.

Before that, what you describe was called 4GLs, Oracle Forms etc. After that it was called "low-code" or "no-code". A more serious effort in the 2000-2010s was Model-Driven-Architecture. All worked for a short while, as they made creating apps for the current technologies easier, but failed when a new technology came out:

• Desktop-based platforms could never create good web apps. The results were too heavyweight or took so much time to migrate, the apps were abandoned and rewritten.
• Web platforms were terrible for mobile, even for mobile web.

Even well-written MDA applications couldn't handle the usage patterns of new platforms like web or mobile. Those usage patterns meant the data and processes changed as well.

Martin Fowler was skeptical about MDA and the CASE tools that predated it as far back as 2004

It is not going to be obsolete, especially when deploying systems at a large scale, you can use AI to assist you in the process but you as a developer need to have knowledge on how to scale the app, how to maintain the system generated by AI.

Everyone can now build and be a vibe coder and use AI to build slop shit but not everyone can maintain and scale applications to many users or to simply deploy an app and have security standards on it.

Manik, In 1980 I read a book Application Development without Programmers, by James Martin, which predicted that users would be able to create their own programs using 4th Generation Languges, and that programmers would be obsolete by 1990. Most of the programmers today have been born aftrr 1990.

Well following that logic, in only 66 years -- 1903 to 1969 -- we went from a 120-foot flight to a roundtrip to the moon & back (over 500K miles). Therefore by 2035 we will have the capability of travelling over 11 trillion miles through space -- halfway to Alpha Centauri.

As for regulations: have you ever worked in a regulated environment? I have. I spent 4 years in the V&V (QA/test) department of a nuclear-power software supplier. We examined every single line of code the developers wrote, and we filed a ticket for anything that didn't conform to the mandated coding standards. We then tested every single line and every single branch -- 100% coverage of each -- plus a minimum 85% coverage for a higher-tier metric called LCSAJ (linear code sequence and jump). We did all that -- and more -- because it was mandated by 10CFR. Don't expect those regulations to change anytime soon.

I don't think coding is going away, but it's definitely changing. AI can already handle a lot of repetitive coding and even generate working apps for simple projects. The bigger challenge is still figuring out what to build, designing the system, handling edge cases, and maintaining it over time. Even if we eventually describe software with diagrams or natural language, someone will still need to verify the output and make decisions that AI can't reliably make on its own. It feels more like the role of developers is evolving rather than disappearing, similar to how high-level languages reduced the need for assembly without replacing programmers.

Depends on what excatly you mean with "coding". Yeah, maybe writing code in languages like TypeScript, Pyhton, C ... may become more and more obsolete with emerging AI. But isn't creating those state diagrams -- especially at such a detailed level, so they define all aspects of you application -- also a form of coding? It's just a different language ...

Same as a simple C++ loop for (int = 0; i < x; i++) { ... } will be compiled to some assembler code, shifting around the data in registers, making the necessary ADD, CMP or JMP calls ... also a state diagram with its states, conditions and transistions will be "compiled" to some sort of code. And in the end of the day, at least nowadays it will always go all the way down to some sort of assembler, because that's the only language our current processors are able to understand ...

I don't think coding will become obsolete, just optional for many tasks. There will always be people who want to hand‑code, the same way there are people who still use analog calculators.

If I enjoy coding, then I'm going to code. If I have no idea how to code, or don't want to learn it, then I will probably rely more on AI.

I don't think we are far away from this. We've made an enormous progress in few years. I don't even want to imagine what the AI will be capable of by the end of the decade.

As for regulations: that's not true at all. Do you think people read the assembly code after compiling C/C++ code? Of course they don't. They simply put trust in the compiler. Mostly they don't even read the source code, millions loc is too much. They simply trust in tests or other tools provided by someone else. I don't think this will be any different with the AI. Humanity, governments and regulations just need time to adjust. And I don't even think this will take long. In fact in the military sector for example, AI already is extensively used for coding. Whoever sleeps, loses.

Short answer: No.

Coding is a way to precisely describe to the compiler what it should do.
As you said, natural language is not precise enough to describe that, therefore we have the rigid, structured programming languages that essentially tell the compiler "Do this. Exactly this." - and the compiler, with zero creativity, following a similarly strict set of rules, does exactly as you told it (including any mistakes/bugs you included in your instructions).

Now let's switch the view to AI. AI, by both design and necessity, is not following a strict rule-set. Its outputs are not precisely predictable. Even if we pretend it could correctly "guess" at your intentions you still have the first problem: Precisely stating them.
Can you precisely "draw" your desired states and outcomes in a state diagram? "Click button to open window" still sounds easy, how about "Take this list of photos and sort them"... sort them how? Alphabetically? By Date? By image content? What exactly does take mean? You are giving your AI-Djinn a lot of wiggle-room here and, even if it is benign and tries to interpret your wish as correctly as possible - natural language is simply not precise enough.
So we once again return to an agreed-upon, formalised language, that precisely describes what you want and do not want, and we're back to coding. (And this argument entirely ignores how annoying it would be to be forced to express yourself purely in state-diagrams... personally I don't like playing charades with the compiler)

Therefore by 2035 we will have the capability of travelling over 11 trillion miles through space -- halfway to Alpha Centauri.

There are fundamental physical constraints here. There are no such constraints for the AI. Or at least none are known yet. That's one thing.

Secondly, I have no idea how you calculated this, or how does this even refer to what I said. I said I don't know what will happen in the future. But AI is a very new and very rapidly growing area. The fact that it achieved so much in say 3 years, and also the fact that we don't see any inherent obstacles for growth, means that it will grow for quite some time.

And also, seriously, the static analysis you refer to in the second paragraph is already the place it outperforms humans. I use it every day to statically analyze my code. And it catches subtle mistakes that I make all the time. Its already better than most of the programmers in localized static analysis.

We examined every single line of code the developers wrote

That's cool. And you also examined every single line of the generated machine code? Or did you trust that the compiler generates correct code? Similarly to how you would trust the AI? Don't answer, I know you didn't.

Don't expect those regulations to change anytime soon.

Those regulations change regularly. The last major change to 10CFR happened in March 2026, 3 months ago. So I don't know what you are talking about.

And also, it really doesn't matter if there is one or two areas with the archaic structure, 100 years behind the progress. That's an irrelevant nieche, from the developer's point of view. For example, did you know that there are still serious businesses running C64? There is one famous example in Poland, where C64 runs non-stop for decades now. Noone cares. And I would bitch-slap anyone suggesting that learning C64 for any other reason than history/archeology is ok. Unfortunately I think that the entire programming will share that fate eventually, and I think sooner than later.

computers are very very fast at being very very dumb. AI needs "training"... by humans. Eventually it might be able to guess at a reasonable solution using old code written by humans. But if coders stop coding then AI wouldn't have any new code to "learn" from. It only makes statistical guesses so I think it needs very large datasets to "train" on.

It remains to be seen if the future will be different somehow... I'm not sure that computers will ever stop being very very dumb at their core.

True. But even in different contexts (professional or business), the choice of tools or technologies is still based on scope and preference. I think there will always be at least a niche for hand coding.

Yeah, there also are people that still code punched cards. But I don't think this question is about hobbies or oddities.

Do you think "jobs" as we have today with a large company and a hierarchy working on the same project exist in the future or will people earn money by running individual (or smaller) businesses/contracts only?

I've found Claude Code and Codex like tools is like having a team of nerd rookies that just graduated from top notch universities, lot of knowledge but zero engineering experience. BUT, if they are not managed by an expert software architect (human) they tend to write extremely bloated code going up and down with trail error strategy, same way as an unmanaged and unsupervised super nerd rookie would do. In the long run, I think yes, coders are doomed if they don't move to management positions and software architecture and engineering. However, with current AI pace, probably in 5 years more that position would be at risk too.

The capabilities of AI coding tools, impressive as they are, have still been vastly overstated. We are a very, very long way from AI being able to generate the millions of lines of code needed for a large scale system like an Operating System, an Enterprise Resource Manager, a Warehouse Management System, a Massive Multiplayer Game site, etc.

Another barrier is the fact that many software applications are subject to government regulatory requirements. Examples in the US include medical devices, transportation, nuclear power, banking or payment processing. The regulations simply will not allow the code for those systems to be generated by AI unless the vendors have a human review every single line for compliance, and an independent verifier exhaustively test the same... which largely defeats the purpose of using AI.

"what cultural and social changes are made": avoid Kindergarten: https://sites.google.com/view/davidlaczko/home/short-stories/kindergarten

Policy: Generative AI (e.g., ChatGPT) is banned

I am inclined to think that the days of the journeyman coder are probably coming to an end.

AI like Mythos is now smart enough to tear holes in human developed code that is supposed to be military grade secure (and human verified as such). AI will be a tool that allows system designers to leverage AI coding that turns requirement specifications into more reliable code.

Perhaps Nassi-Schneiderman diagrams (or a similar notation) are due a come-back (nasty spiderman diagrams as some would call them). The other possibility is that with AI code generation mathematical proof specification languages like Z & VDM could gain ascendency at least for mission critical code.

There will always be jobs for the best systems analysts who can understand the users domain and translate that into what computers/AI can do to supervise the AI though. AI has a tendency to hallucinate solutions when given a free reign and will invent case law to support its line of reasoning.