My favorite part about gushing about the new age of AI development is how oblivious the authors always seem to be about the next phase.
It's always some variation of: the LLMs are incredible and gaining logical and reasoning ability at an astounding rate, but fortunately they will always need me to sit here to hit the "Enter" key.
until 1) software is saturated or 2) llms or their creators will take accountability for any issues its code causes i think there will continuously be the next phase
Tbh i dont see software demand being saturated anytime soon (which is why im bullish SWEs, and we're seeing both demand and compensation increase as of late), we keep moving up abstractions and increasing ambition and scope for new projects.
If all the companies use {frontier model} then your competitive advantage is what you have that is not legible to the model
There are a bajillion apps on the app store that get little to no recognition. Most software startups fail to find a foothold and go under. An uncountable number of FOSS projects have been seen by at most a handful of people, and even fewer who have used/cloned the code.
Well I mean - I don’t think it’s so much that you’re needed to hit “enter” - anybody who’s used an agent on full auto has already experienced that lack of need.
What is needed is a human at the beginning, to decide what needs to be done - and a human at the end, to decide whether it’s finished.
For a while at least, specialized knowledge will be a huge asset, if not necessarily strictly necessary, for the beginning and end.
Possibly even that might pass away - but I m a little more skeptical there, because a fully automated proceeds of choosing what to work on is really a whole different world altogether. It’s not just one where humans aren’t required to do menial tasks, it’s one where humanity isn’t required period - at which case, what’s the point of any of this.
I don’t see us getting to that point within my lifetime.
> fortunately they will always need me to sit here to hit the "Enter" key.
Nothing fortunate about it. That's the part I hate the most about today's LLMs. That you still have to do QA for them regarding the design. I should be able to arrive at cool result with 5% of current steering effort, if any. I still feel like a bottleneck between nothing and cool stuff existing. Unfortunately it's unavoidable until they read my mind and know what cool means for me.
While that sounds a little doomerish and probably for good reason, there is the obvious alternative.
Dont change. Software developers do not have to change anything they control their means of production and we dont HAVE to use any of this stuff. IF you are saying that there will be non-engineers willing to step in to hit enter (as in the business/product people) and LLMs will be good enough to do with it, Im ok with that.
I encourage business people to go down that path, especially since I have no organizational responsibility to help them. So if they run into even one little issue, I wont look at their code, I wont let it into a repo and I can just walk away.
There will be people who cross over and will want to help the people using LLMs and some will be technical. But it wont be in large numbers and it will be based on extraction and not pay very well at all. It will pay less every day as the LLM usage goes up to the CTOs desk.
More power to them! At the end of the day someone WILL have to hit enter, and it wont be me in some low paid slopshop.
What Im saying is the industrys future and mine are not intertwined. The tech industry is of no interest to me as a software engineer. They have jumped their own fake shark.
Ok but... You aren't planning to get paid to be a software engineer? Or you are? Surely this is the way in which the futures are entwined, for most of us?
> And yet, how also we have changed, and continue to adapt to the new ways of working.
Assuming this trajectory continues, the best case is the only adapting which we'll need is to sit in the assembly line chair: wages reduce to that of a skilled factory worker, little to no leverage in labor market, and economic precarity.
>the new era of swe will demand more and more long term business + product direction and relationship building skills
Why do you assume this to be the case? Look at your own company culture--has it not become increasingly siloed as people look to their own agents for most questions, tasks? What relationship skills are you building while talking to your chatbot all day?
>the trajectory leads us to all becoming mini-ceos directing our company of agents
Only if you completely misunderstand how business and economics works. The price of code is falling because it's turning into a commodity. The amount of people needed to produce it is drastically decreasing.
We consume more food than ever, but the % of people who work to produce food is far smaller than it was 150 years ago.
> has it not become increasingly siloed as people look to their own agents for most questions, tasks
doesn't feel like it, collaboration is more important than ever for high level direction even if nobody pair programs anymore. relationships feel more important than ever especially customer ones. theres no RL environment for competing in business
> We consume more food than ever, but the % of people who work to produce food is far smaller than it was 150 years ago.
Yes, because there is a limit to human desire for food, as we tend to get full. is there a limit to humanity's appetite for software? I would posit probably not, because software is infinitely abstractable and can be applied to any industry and domain
the price of code is falling but the value of differentiated ideas and market alpha is rising. human attention is scarce - as the mini-ceo with your army of agents, how will you compete with the other mini-ceos with their army of agents that can produce software at a whim?
> We consume more food than ever, but the % of people who work to produce food is far smaller than it was 150 years ago.
I’m not sure this analogy works. In fact I’d argue it’s the opposite. Food is an example of something that is relatively stable in terms of consumption per capita. There’s only so much food a person can eat, so if you make it easier to produce, then jobs producing it go down, as we’ve seen. But software isn’t like that at all. There’s essentially no end to how much software we can consume, and so as it becomes cheaper to produce, we’ll likely produce it in even larger quantities. Again, as we’ve seen.
> We consume more food than ever, but the % of people who work to produce food is far smaller than it was 150 years ago.
You need to normalize that with world population, or a bigger population and a smaller percentage could be the same number of people.
Of course it is still a bit simple, because the distribution will not be uniform per country.
Then we need to know why 150 and see how other numbers look like, ideally per region. My guess is industrial revolution but it is mixed with different timings and political events in different places.
Regarding
> What relationship skills are you building while talking to your chatbot all day?
Obviously talking to the client, or does your llm power itself on and starts working without any input from you?
I've actually been noticing this with a bunch of staff+ engineers at my current workplace. Previously what could be done with a TL plus 2 or 3 juniors can be done by a single person. I'm seeing some really sophisticated internal tools being built to help internal users with issues that previously required oncall engineer time.
> the new era of swe will demand more and more long term business + product direction and relationship building skills
This is what agile software development was always meant to be in fairness. It was the business people wanting to infantilise technical staff that created all the layers and walls away from users.
Maybe so, but what I have found in my career long before AI is that a significant of percentage of biz people lack the capability to think holistically or end-to-end. Product people are a little better on average. But at the end of the day it is senior technical folks with product and UX sensibilities who are able to determine what most stakeholders actually want. I find the infantilisation to be going the other way.
Enter key? Just talk to the LLM, no enter key needed ;-)
Now anyone who can talk can vibe code apps.
Personally I vibe code apps while driving. I still have to test them once no longer in motion, so im thinking maybe humans are still needed for testing but in a few years or more maybe not?
How so? For safety its the same as hands free phone conversations in the car which is perfectly legal. Also, it's much safer then looking/using phones (that's illegal) while driving, which millions unfortunately do now.
Also, talking and listening to AI for directions is safer where the Ai is like an intelligent person. An example the AI giving you contextual directions like that new mechanic you are looking for is next to your favorite McDonalds in XYZ city or town (provides context to something you're familiar with so you just drive to a familiar place).
>How so? For safety its the same as hands free phone conversations in the car which is perfectly legal.
The amount of brain power most people use while having a passing conversation is << than the brain power people use while trying to solve a problem, or reason about engineering abstractions.
I don't think this is a responsible thing to be doing unless you've completely relinquished all control of how things are built/designed/implemented to the AI.
Just because you __can__ does not mean you __should__.
You think programming still requires a lot of brain power? Definitely not as much as anyone who can talk can create software, iPhone apps, etc, etc.
Recently on a long drive I was working on an AI agent network concept talking it through and after each push to it's repo I would discuss more and prompt it to make more updates. On Wednesday at a bar with 2 unemployed software dev friends they were testing out my dating app Ive been working on in the car/in general. They provided me feedback on the dating app and boom GPT made the updates in less then minute.
"Just because you can___doesn't mean you should," I agree with you, as I wish this didnt exist, but it does and times always change. I'm personally embracing this new change!
>You think programming still requires a lot of brain power?
Maybe not for you.
>"Just because you can___doesn't mean you should," I agree with you, as I wish this didnt exist, but it does and times always change. I'm personally embracing this new change!
You can embrace AI agents all you want, but don't risk other peoples lives by distracting yourself while driving.
>The amount of brain power most people use while having a passing conversation is << than the brain power people use while trying to solve a problem, or reason about engineering abstractions.
Ok got it never have a handsfree phone conversation while driving if it involves a lot of brainpower. Which is super subjective and id say talk to chatGPT to create an idea of yours and see if your mind changes on this topic ;). It sounds like you have not talked to conversational AI to create software yet. cheers
Until either the enter key wears out and breaks, or we get pervasive sandboxes that act according to the users interests. It seems like Windows Vista UAC all over again for a new generation.
We'll keep doing it for as long as those making the blog posts keep making the same sort of blog posts we have seen on this cycle several times before. Slop begets slop, human or automaton made or a mix, and always has done.
I maintain a browser extension to nuke the endless stream of boring AI noise articles: https://github.com/ColdPie1/enough-ai-on-hn/ Sometimes it misses articles with less clear titles, like this one, and you gotta nuke 'em manually.
It's always some variation of: the LLMs are incredible and gaining logical and reasoning ability at an astounding rate, but fortunately they will always need me to sit here to hit the "Enter" key.
I envy their optimism.