Try that with photography, painting, or music: "AI music has promoted musicians to architects".
The idea that rationalizing a job more (relying on metrics, measurements, statistics, predictions) will lead to better labor outcomes is foolish. We are only paid well as devs because we have comprehensible jobs (ones that require specialized knowledge and understanding).
We have over 200 years of history regarding what happens to labor when their industry becomes more rationalized and it has never ever benefitted labor once in the history of capitalism.
The medical industry achieved some protection through regulatory capture, licensure, and throttling education. Doctors have very high salaries now but there's a shortage and they're overworked. Innovation is also throttled in that space as a result. Socialized systems aren't faring any better, the doctors I know in Canada and Japan are just as overworked, have to rush cancer patients through 5 minute appointments, and are paid far less.
There is no magic bullet for making a large profession sufficiently protected by mystification (anti-rationalization?).
in the age of AI, the helpdesk/call center is really not different from that of a software engineer. The only diff is Helpdesk operator talks to a human. SWE are now just talking to LLM.
the job is perfectly mobile, can be offshored, outsourced, replaced, and million other ways to destroy any attempt to unionize.
unless, the society-wide labor unites as a single voting block for more redistribution from capital towards labor and universal federal ad-valorem wealth tax, instead of income tax
I don't think that's going to work. There's no barrier to entry for vibe coding. Your union would have to include everyone in the world.
The interesting thing is that this exact thing happened before in the software industry. It's what killed the "old" software giants of the 90s. Only really microsoft was able to pivot and recover.
"Big Tech" became big tech by bucking that trend and being much more empowering to "low level" employees. Allowing talented people to build good things. They ditched the measuring productivity with "lines of code" and tickets closed and all that nonsense that was common place before that.
The current batch of tech CEO are going down the same path of morons like Jack Welch and we're all going to pay the price for it. I guess, if I were them, it's a reasonable choice. Despite leaving a wake of carnage and turning great companies into shells of themselves, they'll be billionaires before anyone figures out how utterly fucking stupid they are.
It's just very odd how, as an industry, we learn nothing from the past. Not at the executive level and not at the worker level. It reeks of arrogance at all levels.
It’s just creating your own team of little robots working for your own personal “department”.
Though what I am seeing is that increasingly I am “setting up teams of agents” which each have their own “lead” and I just talk to these “leads” not individual agent “developers” . Effectively becoming a “VP” of engineering of my own private little “consultancy” … maybe next year (or month?) I will manage a bunch of “vps” as the “ceo” of said mini organisation … a product engineer indeed.
More often than not, one person does the majority of this, and does their best to break up the responsibilities. A good CTO does that at the top, and it flows down through the ranks depending on how big the org is. Perhaps it's just a CTO and 5 engineers, perhaps it's a CTO with 5 divisions, each with 10 groups, with many teams.
Flatter structures of engaged members works best from what I've seen, but you need to have a critical mass of people not only willing to take initiative, but empowered by leadership.
What is happening is no different when developers got promoted into team lead of the offshoring team.
Which also didn't require that many team leads onsite, and off they were, everyone else that wasn't leading any offshoring team.
It was mostly because in some companies there was a cast of 'architect' who were seen as above coding, and who mostly delivered out of touch UML diagrams, or ill-fitting data-structure. The core of the issue was that even good programmers would rust and get incapable of understanding the ground truth if the organization discourage them to also be in the implementation side.
Here it's not exactly the same thing with AI, since the loop is dramatically shorter, but the (lack of) feedback issue remains.
I really believe that my added value as a programmer was fundamentaly to evaluate technical tradeoff of the different options to solve a business issues, and them to implement them. Now with Claude code I'm infinitely faster at the implementation part, but I'm also eroding my capacity to evaluate those tradeoff.
Programming is a medium, and you also need to work with it from time to time in order to keep some level of competency. And yet in many companies we are basically forbidden to do that.
I feel that we have been given cars, but have also been forbidden to ever walk again; Sure the car is more efficient most of the time, but there's a an almost-vital reason why we have to do jogging to compensate.
Architecture is no more immune to AI than software engineering.
OP is describing somebody using AI poorly, and seems not to have a great handle on the way all jobs are impacted and boundaries being wiped away.
> AI coding tools are tuned to implement your functional requirements and two non-functional requirements, i.e. security and backward compatibility to avoid breaking your app. This invariably leads to:
- Architecture that trades off performance for security and slows down the app.
- Multiple versions of datamodel design with redundant tables, columns and code.
These foibles come and go: early Claude Code refused to let programs crash, mangled all the comments in your codebase, and then one week you find that it insists on ending all functions with `Of` as in `layoutOf()`, `viewOf()`. Finding and mitigating these annoying habits becomes part of the job. Code review becomes articulating with this entity why certain patterns in the code are silly and should be changed.
The author mentions performance and security issues as the domain of “the architect.” For performance, I’ve found that Claude Code has more or less solved this domain, and can profile and optimize code across a variety of stacks and languages using sophisticated, ad-hoc workflows. For security, it’s just an ever-ending nightmare where random habits and sloppy implementation create random vulernabilities, compounded by the fact that Anthropic has added “safeguard filters” whenever you naively say you’re looking for them. Nevertheless, I’m not sure the architecting role as they describe won’t be automated in the near future.
functionally similar
If you go from a (presumably) reasonable load time to one of 15 seconds and it’s not caught before getting merged, you have bigger problems than developer vs. architect and AI coding.
The premise of the post has merit, but the example is a terrible choice.
Is it database requests? RPCs? Is there a function call which is computationally expensive?
Teams then respond by making rules to help make slow things stand out "no RPCs in functions, no synchronous RPCs, only 1 database read per request, don't call this code in a loop", etc.
Caught before merge? Only if you have someone really familiar with all the abstractions _and_ the team's previous failures reviewing it. Otherwise, developers are always finding new ways to write slow code. :)
But, that's why we have feature flags and incremental launches.
If you weren't doing it, you were doing it wrong.
But they don’t teach you that in school. You get how and a bit of what. Maybe a little why in a sales pitch or an aside from a professor who hasn’t worked in industry except in research which is prototyping and we all know how people who only do prototypes are about rigor.
[and your director now is convinced all his stupid ideas are both brilliant and very safe]
Priorities.
Useful question for "analytics" people. If you only collected data for every thousandth customer, would you have less useful insights?
If AI is even remotely capable, we should fire all staff, and hire the same number of people in Nigeria! Diverse workforce is by far the most important goal! And it reduces CO2 emissions from commuting! And we save money!
AI is great at writing code. But it doesn’t know if you are launching the app or have millions of users, which it needs to make the right architecture call to log a timestamp vs email. It is also sycophantic and tries to execute your requirements exactly without pushing back, which means you end up with a slow app that takes 15 seconds to load a static welcome screen.
It is so paranoid about security that it will never write code to accept an unauthenticated payload. Even if you write it yourself, it will rewrite it into an authenticated payload and screw up performance or force users to sign in again on your app.
Just incredible seeing this post at the top next to "Unforced Navier-Stokes singularity, and the Generalized Riemann Hypothesis". We're at the "Japanese soldier who kept fighting 29 years after WW2" stage of AI denialism.
Jesus, this is gonna end up being middle managers all the way down. Get a grip!
Lol, sounds like something only a junior developer or intern would say.
From that perspective, engineers are still engineers, just without the hands-on-keyboard coding. If that's all your job was, hands-on-keyboard coding with no decision making, yeah you're cooked and need to reskill.
Yes, that is how most functional organizations operate.
Architects are a bad meme, along the same lines as scrum masters and product managers. Good software engineers spend more time on design than implementation, and the best are helping their teammates with design, and auditing designs of adjacent teams to ensure integration success. So architecture is an important part of the job, but that doesn't mean that you can automatically declare a role with the title "architect" and make the whole organization do better design.
Architect roles are a bad meme because the architect always ends up knowing less about the system and how it is used than the engineers and support staff, yet they have more influence over the decisions affecting those things. That's the essence of an organizational anti-pattern/bad-meme any time it shows up: influence over resources anti-correlated with decision quality. Additionally, architects often report to non-technical folks fairly high up the chain. Those people are bad at screening for such a technical role, sometimes even making the mistake of replacing an architect with an external hire.
And that’s why junior engineers are fucked at companies that can afford to send all their proprietary information to Anthropic. But since they are public now, they will violate that trust as soon as major shareholders learn to word it so that it sounds like a stranded asset that any responsible C Corp would be expected to rescue. And then everybody’s fucked.
After getting complaints, I started getting narrower and narrower in my ticket writing, until I realized many newer software engineers literally want tickets like "modify function x_y_z() in python_file.py so that it also returns..."
In the distant past of a couple of years ago, I saw and gave tickets with higher ambiguity and opportunity for ownership and cross-team collaboration. There was breathing room to take that kind of time.
This year, however, as we're all aware, the expectation around speed is much higher. When more spinning plates are added, I want them to be more defined so they can be handled quickly. To be clear, though, I much prefer the former scenario.
Now with AI coding, it pisses me off even more.
You aren't an outlier exactly, but there are way more devs than you are probably aware who are just told what to build and then have to go build it. So they don't have or never developed the consulting, planning, and architecture skills and experience.
The more our clients tend towards wanting a process like that, and the less they treat us like valued expert software consultants that can help shape their solution for the better, the worse the outcomes usually are.
There's really no sense in handing off the design to a dev to implement, who just copy pastes it into Claude's prompt. The "Architect" now needs to be fully responsible for what goes into production, instead of being able to blame some poor dev who faithfully implemented a flawed architecture design.
AI is flattening architecture and development into a single role.