90% of the problem an architect has is being able to understand and thus express the requirement and limitations of a system, and understanding how it interacts with everything else.

I.e writing the prompt, understanding the answers, pushing back etc.

90% of the problem an architect has are people.

This!

Unless AI is introduced as a regular coworker and stakeholder wants to communicate with it regularly, I don't see this changing anytime soon.

Reverse engineering stuff when non-cooperative stakeholders dominate the project has its limits too and requires "god mod" type of access to internal infrastructure, which is not something anybody gets.

That's because a large percentage of "architects" aren't really all that great. We interviewed a candidate that didn't knew much of anything, in terms of actually operating IT infrastructure, but in their mind that didn't really matter because they where looking for more of an architects role and didn't want to touch things like terminals, YAML, databases and all that stuff. They completely serious just sat there and told us that they really just wanted to work in diagram tools and maybe Excel....

Architects are like managers, it's way harder than people imagine and very few people can actually do the work.

Yep. It's indeed like "management", where people expect to just slide into it as a reward for staying at a company for a few extra years.

Also I hate that "architect" is used as a synonym of "cloud architect". There is much more to software architecture than cloud.

Precisely. I noticed that a previous intern of mine is now an "Enterprise Architect". He's a smart dude, no doubt about it, but from zero to architect in 3.5 years? There's no way this person has the experience to be an architect. That is a promotion because the company either needed someone with that title, or because he was "paid off" to stay onboard.

> We interviewed a candidate ... didn't want to touch things like terminals ... they where looking for more of an architects role

I don't know whether you were actually looking for an architect? There are different types of architects. For example, you have got enterprise architects that indeed will never touch yaml, you have got solution architects who have a more narrow focus, and you have got engineers with a plate of overbearing work and team responsibilities. The latter are better called lead engineer. In my experience, being a good (lead) engineer doesn't make one a good architect, but companies try to make their job posting more sexy by titling it with "architect". One would imho do better by taking lead engineers seriously just in their own right.

Architects in general need to be very skilled in abstract reasoning, information processing & conceptual thinking. However, the people hiring for an "architect" often look for a 2x/nx engineer, who is able to code x widgets per hour. That is a stupid mismatch.

I would agree however that someone without previous practical experience would be rather unsuitable, especially "below" enterprise architect level.

This is what wishful thinking looks like. The author is probably proud of their architecting skill so they think it's irreplaceable. If they were good at, say, optimization, they would think optimization is irreplaceable.

I think that's just in the nature of these tools. They are better at doing things you can't do (most of the things), but worse at the things you can do (very few things).

Ex: If you're a lazy typist like most, then a code assistant can speed you up significantly, when you use it as an autocomplete plus. But if you're a very practiced vim user and your fingers fly over the keyboard, or a wizard lisp hacker who uses structural editing, then a code assistant slows you down or distracts you even.

Or as Marc Andreessen said, being a VC is the last job AI's will be able to replace :)))

> Andreessen said that venture capital might be one of the few jobs that will survive the rise of AI automation. He said this was partly because the job required several “intangible” skills and was more of an art than a science.

https://fortune.com/article/mark-andreessen-venture-capitali...

> the job required several “intangible” skills and was more of an art than a science.

I've seen a lot more ai-generated art than ai-generated science.

No, AI has proven quite adept at generating self-aggrandizing bullshit.

Architects what are they?

Ohhh, you mean power point writers. Sorry, lost you for a minute there.

I’d frame it a bit differently. LLMs are pretty good at generating the midwit solution to problems because that is the bulk of the available training corpus. It is a generic “best practices” generator. You would expect it to be better than a third of human architects almost by definition.

On the other hand, they are pretty poor at reasoning from first principles to solve problems that are far outside their training corpus. In some domains, like performance-sensitive platforms, the midwit solution is usually the wrong one and you need highly skilled people to do the design work using context and knowledge that isn’t always available to LLMs. You could probably use an LLM to design a database kernel but it will be a relatively naive one because the training data isn’t available to do anything close to the state-of-the-art.

> Yet we never 'see' this in the article. It just restates it a few times without providing any proof.

I'm honestly shocked by the number of upvotes this article has on Hacker News. It's extremely low quality. It's obviously written with ChatGPT. The tells are:

(1) Incorrect technology "hype cycle". It shows "Trigger, Disillusionment, Englightnment Productivity". It's missing the very important "Inflated Expectations".

(2) Too many pauses that disrupt the flow of ideas:

- Lots of em-dashes. ChatGPT loves to break up sentences with em-dashes.

- Lots of short sentences to sound pithy and profound. Example: "The executives get excited. The consultants circle like sharks. PowerPoint decks multiply. Budgets shift."

(3) "It isn't just X, it's X+1", where X is a normal descriptor, where X+1 is a more emphatic rephrasing of X. ChatGPT uses this construct a lot. Here are some from the article:

- "What actually happens isn't replacement, it's transformation"

- "For [...] disposable marketing sites, this doesn't matter. For systems that need to evolve over years, it's catastrophic."

Similarly, "It's not X, it's inverse-X", resulting in the same repetitive phrasing:

- "The NoCode movement didn't eliminate developers; it created NoCode specialists and backend integrators."

- "The cloud didn't eliminate system administrators; it transformed them into DevOps engineers"

- "The most valuable skill in software isn't writing code, it's architecting systems."

- "The result wasn't fewer developers—it was the birth of "NoCode specialists""

- "The sysadmins weren't eliminated; they were reborn as DevOps engineers"

- "the work didn't disappear; it evolved into infrastructure-as-code,"

- "the technology doesn't replace the skill, it elevates it to a higher level of abstraction."

- "code is not an asset—it's a liability."

---------

I wish people stopped using ChatGPT. Every article is written in the same wordy, try-to-hard-to-sound-profound, ChatGPT mannerisms.

Nobody writes in their own voice anymore.