AI and globalisation are shaking up software developers' world

Whenever one reads articles like this, it's a bit of a consolation to remember that the same factors are shaking up the journalists' world much more and much sooner. This article, like many of its kind, could've easily been written by AI or by a contractor in a cheap country; in fact, this time it just may have been.

My friend dubbed this effect "AIrony"

I think Cursor is great, it's really boosted what I can do.

I think it's made me 3 -> 5x more valuable.

Now, maybe it's done the same for other people as well, but I think my co-efficient was high, and I continue to insist that I've never worked anywhere that had enough resource focused on software development. The demand has always massively exceeded the organisations ability to fund supply.

And, as many have noted lower costs can also boost demand.

Finally, off shore devs are on a longer cycle time for many reasons. If cost onshore are reduced then that cycle time will crush offshore competitive advantage. I don't see any prospect of me getting work from India or Indonesia, but I do see AI as a driver that will reduce work leaving the UK for other places.

the execs are just salivating to commodotize us. my experience with outsourcing tells me we are fine.

I feel like creating software applications is not what leads to high salaries as a software developer.

Its creating software applications and then signing up to support said software codebase for 100K+ users and agreeing to pager duty and triage management for it for more than a 5-10 years.

Perhaps AI purveyors should focus on automating customer/bizdev support by creating frameworks/workflows for AI pager duty and AI triage management.

This again? What year is it?

We need to move past this type of thinking. Most professional services could be automated and yet they are not for a number of reasons. For developers, the unescapable complexity that keeps creeping into projects if you don't know what you are doing will keep us employed.

People also dont realize that this will kill SAAS margins. If anyone can quickly build working software, the whole moat is gone.

If you dont think this will not happen, you have not used cursor.sh with sonnet 3.5. I am working on a startup that is effectively a copy of existing products. My flow goes:

1. Screen shot competitor UX

2. Drop screen shot into cursor.sh, ask it to code it (there will be issues)

3. Screen shot individual components in the above screenshot

4. Have cursor.sh fix each component one by one

5. Go back through and tell it each issue by highlighting the code

6. Iterate a bit

7. 90% of the UI has been copied in one day, with a working backend.

This is with a next.js/supabase app deployed on vercel. For complex B2B saas.

Its over. In a year the models will get so much better

We are going to see a massive creativity expansion with the many, many people who will have access to amateur level coding. Think of it like the expansion of visual images once the first personal camera was widely adopted. There are still professional photographers, but they are now dwarfed by the number of amateurs (everyone with a modern phone).

Except this analogy breaks down because photography is but a sliver of what is possible with ubiquitous programmability. It's going to be more, different, and somewhat unexpected.

Get ready!

I think we can all agree on those points but this is a very superficial analysis. Now that the barriers to coding are lowered, will that increase or decrease supply, what types of new tasks (or job titles) will AI create?

Everyone is getting accelerated learning and task completion, but what does that mean for the market? I think there are more fundemental things to this change that are needed to be fleshed out. Pre LLM programmers and post LLM programmers will be very different.

In my observation, the software written even now--by people who are bothered to both learn the language and put their brain to work--are of potato quality. Brace ourselves for a world where the authors don't have even the same qualification producing even worse: autocomplete-driven development. A world of zero accountability, responsibility, and traceability.

> engineers helped build Teams, Microsoft’s video-streaming service, as well as designing chips and software for “connected cars”,

Ah yes, software famously loved by users. Microsoft teams and in-car software.

In the second chart "Software developer median salary", the bar for Germany is very likely wrong: I am very sure that the wages are quite a lot lower (rather in the 5-figure range instead of low 6-figure range).

This might help with building new software. But large part of the job is maintenance and changes. How long will next outage take to understand without people with deep understanding of the code.

What I don't get is, how on earth developers in the USA jumped on the boat of WFH.

Isn't it obvious that if a job can be done from anywhere that would mean that you are opening yourself to world wide competition when subsidizing your company with your own facilities?

Was it hubris? Do Americans believed that only they can do this coding stuff? Not only that the job market was opened to global competition, it allowed tech and capital transfer. On the grand scheme its great, now everyone does all these things but from workers perspective its very bad idea.

It has been quite some time to see some outstanding American innovation. OpenAI and the generative AI was one but it didn't took much time for China to leap ahead. Maybe its nationalism, maybe its the language barrier but I see a lot of Asian innovations that don't catch much attention in the Anglosphere. American techies pumping Sora on Twitter over some demos when Chinese are having it open to use by the public was amusing.

There's this narrative on Twitter and elsewhere about how EU over regulated tech and fell behind US but this whole narrative is based on dismissing the Asian innovations. You have to omit China and other Asian countries from the graphs that show Europe is falling behind so to claim that free market and libertarianism is driving innovation and growth, because if you include Asia you end up claiming that its communism and centralized economies are actually the big success stories.

not sure where they got that info for the "software developer median salary" but pretty sure thats not the median salary for software developers in the Netherlands. It's about half that.

soo… just like every other job, they’re getting 3rd country workers to work for pennies and trying to pretend it’s a fair thing?

This is a bit sensationalist of course. But the notion that AIs are enhancing developer productivity is real. I use LLMs for different tasks with coding. Is it perfect or amazing? No. Does it save me time. Yes. Document this code; write me test for x. Cover these cases as well. Give me some quick feedback on what I just wrote. Did I overlook anything? Etc. If you look at my chatgpt history, I'm constantly interrogating it about whatever I'm working on. That saves me a lot of time. It's a productivity enhancer.

Of course the history of computer science is that we've had a continuous stream of productivity enhancing things that helped us become more productive. Whenever that happens, what doesn't happen is a decline in demand for programmers. As what we do becomes cheaper or easier, people want more of it. So far, there's little sign of demand for programmers declining. If anything, it's now so easy that lots of non programmers start getting their hands dirty as well. And why not? That just frees me up to do a bit higher level, more interesting stuff. A lot of programming work is tedious and repetitive. Getting some help with that is very welcome as far as I'm concerned. That just means I can build whatever I have in my head faster. So, it costs less for whomever is buying that. So, they might get back for some more; or raise the ambition level a bit. It's not a zero sum game.

The second thing that the article brought up (outsourcing) is an interesting one. Because they missed one important thing that LLMs also do: which is translating. Outsourcing programming work has been historically hard because of language barriers. Miscommunication due to people not speaking each other's language very well is a big challenge. You can make it work but it takes a bit of effort. I've worked with people that I can barely understand or that barely understand me. It's tedious and frustrating if you can't understand each other.

That language barrier is melting away as translating between languages gets easier. We're not quite at the babelfish level just yet but you can get pretty close if you have some patience. Meetings can get transcribed & subtitled (imperfectly) and those transcriptions can be translated. Anything in text form is easy to translate into whatever. In short, we still have cultural barriers but the language barriers are becoming less of a hurdle. The quality and speed of the translations keeps getting better. Real time conversation translations are basically no longer science fiction. And the quality of the translations is more than good enough for a lot of things now. Certainly better than two non native speakers trying to speak English with each other (I'm not a native speaker).

The value of hiring locally is mainly in that you can communicate with people directly. And that's just hard if you don't speak the same language. But that's not going to be as big of a deal as it used to be. The value of being local is still there but it's worth less.

The two added together means a vast increase in the number of programmers. But also in the number of customers. There are a lot of growth economies in the world. And they are all going to need programming work done. And they will want to hire the best they can afford and won't be hiring locally necessarily. We're not talking impoverished third world countries here either.

So, I'm not that worried. The opposite actually.