> Agents weaken the social fabric of engineering teams by reducing interpersonal communication, increasing self-reliance, and hindering skill expression.
I have a feeling that this is true and it's a bit sad.
What's been driving me nuts recently is when coworkers send me clearly AI-generated blocks of text in otherwise-human conversations. Like I'll ask in a chat window "hey, Jim, what's the difference between these two deployment pipelines," and I'll get back a block of text with markdown and headings in the style of our team's AI, and I get this sad feeling where I thought I was going to be having a conversation with a colleague, but instead my colleague was just playing as a chinese room.
There's a particular sting to being passed raw, unfiltered output from the LLM hose. I think you're onto something with the mismatch in expectations (I was expecting a conversation that led to insights we might both share, and the other party was expecting to answer a question quickly without much thought).
I think there's something deeper than that, too. Almost a feeling of betrayal? I cared in part because I thought everyone else cared, too, like we were working towards a common goal. Asking a question in search of shared understanding only to be passed the modern equivalent of a LMGTFY link hurts in a very particular way. I thought they cared about this, too, but they're only trying to get the bag. I get it, times are difficult right now, but it doesn't make it hurt less.
I must be lucky or work in a great company, because I've never seen this, and I think it would be totally unacceptable for a coworker to even think about doing this. What kind of horrible company culture normalizes this kind of behavior?
The most disappointing conclusion of the author's article is that there is no negative business outcome. So there is no incentive from the top to stop the erosion. You don't need teams with high interpersonal communication, you don't need ownership, and your customers don't care how internally alienated your staff could be.
Its inspiring to see folks putting effort into verbalizing the conflicts and unexpected side effects of working with LLMs. Our feelings really do matter at the end of the day.
But I dont know if they matter to the employers who can only see LLMs as "free milk", and developers as the "milking machine".
Oh god that article about Astra is so relatable. I was wondering why it was refusing to write any whitespace and was just spitting out the worst code imaginable, now I know it's not just my harness or prompts. It's truly a horrible coding model.
My work Slack channels are ghost-towns. People are logged in, but they're never saying anything. I often find myself wondering if anyone is even working. PRs have their names on them, even the review comments are posted from _their_ accounts, but it is obvious that Claude wrote everything from the code, the unnecessarily verbose change summary, to their comments. Even their replies to my change requests start with "You're right — …".
I miss the days when the engineering Slack channels had long threads of new and interesting ways to do things, discussions of something we could do next to make work easier or fix long-standing issues. We used to build up our own knowledge of who was good at specific things, what they'd be interested in working on. Now everyone just has their slop-machine spit out every piece of a change for the entire stack without paying any attention to the details.
Fwiw I work at a company where engineers use AI heavily to produce the majority of work outputs but we also have a Slack that's just as lively as ever and plenty of evidence of humanity. Your experience sounds bad and you may be able to find a better one! It's not all bad everywhere.
How established is your team? I can see how teams that existed from the pre-LLM era can maintain their vitality, but would expect newer teams to struggle a lot more. Also, remote or in-person?
It is very possible to use these tools in the wrong way. Companies that understand how to use AI will do better, those that have these 4 problems won't.
- AI should accelerate your understanding of the codebase and technical concepts.
- Skills should ensure a higher quality of code than you could create without AI
- Engineers should stay responsible for the code they product.
- The available time freed up should be used for communication, product and marketing research. The definition of an engineer who just writes code is something that's going away.
Unless something is so simple that the AI can run it autonomously. There are actually many projects like that these days. And in those cases not looking at the code like DHH does is also a viable strategy.
this is exactly it. smart companies create entire platforms with normalized, context-specific conditions and ever-evolving context for teams to operate effectively with AI in. that means artifacts, tests, ADRs, EDDs, etc. the actual engs are in-the-loop for reviewing all of the outputs, validating it against their taste and judgement, and given sufficient time to do so
seeing AI as an individualized tool for single people to labormax with just means throwing spaghetti at a wall and seeing a lot of failures and incidents (much like my old workplace). it's very different from having a driven and iterative system in place with new processes and checks with clear leadership direction
The team thing is killing us right now. Everyone is just kind of silo'd with their agents and they present these entire complex units of work with no input from the rest of the team, and by then it's too late to fix.
The current "slow down" is a back-handed attempt to let the world catch up with the amount of slop that has been generated. The speed and time-to-market bill is piling up and in the next 5 years there will no humans that can understand this mess. So agentic SWE systems, need to change. Either we trust the validation layer at the end, or we take back control and also learn along the way.
> And that’s true: nobody wants to see how John prompts. There is something repulsive and fundamentally unsexy about talking to an agent. It feels fine when you are doing it, but others... ugh, get a chatroom. Their prompts are cringe; their agents are weird. Like looking at someone else’s TikTok feed.
So real. There’s still something fundamentally embarrassing about saying you used AI to do something. This leads to a pet theory of mine that most people use AI far more than they say they do.
I'm a senior software engineer at my company. The AI slop is so bad that it's causing myself code review fatigue. The engineers are also intermediate with 3-4 years of experience.I had one guy report me to his manager because I wasn't letting him merge his PR in because of a critical bug with his code that his precious ai didn't understand.
It never does because however you frame it, AI is always the perfect scapegoat.
"Oh, I'm sorry, the bug was introduced by the AI and it made such a good case for it. Damn you claude, we'll do better next time. I'll add a code-review agent to find regressions."
this seems to be the true killer app of AI - accountability proxies. its interpersonal opium to have a third party that two ostensibly-in-conflict people can sink their negativity into.
its a major reason why military use of AI is so chilling - "look, I explicity ASKED the murderbot not to kill babies when I authorized the napalm strike on the maternity ward! BAD BOT"
My work slack used to have lots of questions. Now very less. That should be celebrated because we don’t have to depend on anyone. But I guess you can paint anything in a bad way.
> That should be celebrated because we don’t have to depend on anyone.
When your work process depends on SAAS you depend on the vendor of that SAAS.
"Everyone" tells me that "noone" wants to use LLM-based code generators that you can run locally because they're so much worse than the newest ones provided by the major LLM manufacturers. If the gossip is true, then it seems like -for the foreseeable future- you're absolutely dependent on one or both of those two SAAS vendors.
Disagree on the slop point. By now the code of frontier models works well, and the need to read it is gone. The Claude word salad problem was pretty much solved with the 5.5 opus release.
The other 3 points made in article make sense though.
> By now the code of frontier models works well, and the need to read it is gone.
Like anything else, it depends on what you're working on. For small personal projects I agree, but in maintaining large scale business applications with millions of lines of code, I am still having to guide the models to reuse existing code and to be more flexible with their data structures to accommodate changing business requirements in the future.
They only see a snapshot of the world the application exists in due to their ephemeral nature, and until that is overcome, they will only be able to learn general principles and not the particular nuances of your codebase that only arise from observing how the users use it on a daily basis.
If you don't need to read the code at all, what are the engineers doing to make the other issues matter? Deskilling and alienation are still about the relationship between engineers and the code.
You still have to orchestrate these things and the wholistic system architectures, pull the slot machine handle many times to get the right thing, and review/verify the working output constantly.
That said, I think we need to be prepared for how a simple prompt like "make it faster" will work in a relatively short timeframe - and where the AI is able to adequately refactor/re-platform/be done in a way where even the architecture becomes obscure to us.
I often use simple prompts like “make it faster” and get back functionally correct commits. When I review them, I often discover that they didn’t do what I really wanted - sometimes Claude made it faster by incurring expensive costs, or doing a migration that would be risky in prod, or making a tradeoff that’s not worth it.
Perhaps I could set up a workflow to handle this without reading the code. I’ve seen people who make Claude fill out arcane templates, and then spend their time reviewing reports and tinkering with loops or whatever. But why would I want to read a TPS report about the code rather than just reading the code?
Again, the assumption that forms the crux of these objections of all of these is that we're going to hit some plateau where we are going to need humans to come back in, reskill, retake over, etc. I don't think most people are prepared for the fact that that's just a midwit problem and the AI will just keep getting better and won't need humans in the loop.
The backlash against agentic coding comes from the split that has already existed among professional software engineers. The ones who are most resistant to it are the ones for whom it has been "just a job", but the ones who love it are the ones who love solving problems. It's now possible for a small team (one person, a few) to solve truly daunting software problems with the help of agentic methods. If you were a 9 to 5er who valued craft but didn't dream about solving problems, AI is a buzz kill.
Im one of the more mid to resistant folks and this has never been just a job to me. I did take immense satisfaction from learning, deeply understanding how things work and the consist and repeatable behavior, which is just not there with ai. Especially the consist/repeatable behavior, which to me computers that behave inconsistently aren't nearly as worthwhile.
> The ones who are most resistant to it are the ones for whom it has been "just a job", but the ones who love it are the ones who love solving problems.
Wow, what a simplistic, uncalled for view of the world. This is just so ignorantly and offensively incorrect.
I really don't think anybody in these "debates", on either side, has a high horse to ride with respect to rhetoric and solemnity.
This in particular is a really silly hill to die on, because literally every debate about software development, from AI to language choice to tabs vs. spaces, eventually settles into a gravity well of asserting that all the good developers agree with your take.
> This in particular is a really silly hill to die on,
Stating that people can't love software development or solving problems because they are resistant to AI taking over their entire job is a silly hill to die on?
I really enjoy what I do, and have done, for the past 20 years. When I first wrote code I literally thought I discovered magic. In many ways, AI has re-triggered some of those feelings!
I also really care about building quality software for our customers.
And it infuriates me to no end when I see huge amounts of objectively incorrect code being written by AI, and the developer behind it stating they have no idea what it's doing, or when people say they can't solve a problem because they are out of AI tokens, etc.
But supposedly these are the people who care about solving problems? Not just that - they are the only people who care about solving problems.
You can choose to take offense if you want, but since I've seen literally the same sentiment expressed about whether people choose Emacs or vi, I have a very hard time taking that offense seriously.
Oh yeah let’s play this game. The ones resistant care about their work and pour long hours getting the details just right. The ones who love it are the ones who just want to close tickets. It’s now possible for a small team (one person, a few) to create an avalanche of slop with the help of agentic methods. If you were a 9 to 5er who valued sprint velocity but never once pondered how a monad is like a burrito back in the day, AI is the All-Mother
In my opinion, you’ve got the situation backwards. The people who see it as “just a job” are the ones who like AI because they just want to get the work done and get paid. The people who genuinely enjoy solving problems, on the other hand, don’t like AI because it takes away the problem-solving part of the job.
I have a feeling that this is true and it's a bit sad.
I think there's something deeper than that, too. Almost a feeling of betrayal? I cared in part because I thought everyone else cared, too, like we were working towards a common goal. Asking a question in search of shared understanding only to be passed the modern equivalent of a LMGTFY link hurts in a very particular way. I thought they cared about this, too, but they're only trying to get the bag. I get it, times are difficult right now, but it doesn't make it hurt less.
But I dont know if they matter to the employers who can only see LLMs as "free milk", and developers as the "milking machine".
I miss the days when the engineering Slack channels had long threads of new and interesting ways to do things, discussions of something we could do next to make work easier or fix long-standing issues. We used to build up our own knowledge of who was good at specific things, what they'd be interested in working on. Now everyone just has their slop-machine spit out every piece of a change for the entire stack without paying any attention to the details.
- AI should accelerate your understanding of the codebase and technical concepts. - Skills should ensure a higher quality of code than you could create without AI - Engineers should stay responsible for the code they product. - The available time freed up should be used for communication, product and marketing research. The definition of an engineer who just writes code is something that's going away.
Unless something is so simple that the AI can run it autonomously. There are actually many projects like that these days. And in those cases not looking at the code like DHH does is also a viable strategy.
seeing AI as an individualized tool for single people to labormax with just means throwing spaghetti at a wall and seeing a lot of failures and incidents (much like my old workplace). it's very different from having a driven and iterative system in place with new processes and checks with clear leadership direction
So real. There’s still something fundamentally embarrassing about saying you used AI to do something. This leads to a pet theory of mine that most people use AI far more than they say they do.
"Oh, I'm sorry, the bug was introduced by the AI and it made such a good case for it. Damn you claude, we'll do better next time. I'll add a code-review agent to find regressions."
its a major reason why military use of AI is so chilling - "look, I explicity ASKED the murderbot not to kill babies when I authorized the napalm strike on the maternity ward! BAD BOT"
When your work process depends on SAAS you depend on the vendor of that SAAS.
"Everyone" tells me that "noone" wants to use LLM-based code generators that you can run locally because they're so much worse than the newest ones provided by the major LLM manufacturers. If the gossip is true, then it seems like -for the foreseeable future- you're absolutely dependent on one or both of those two SAAS vendors.
The other 3 points made in article make sense though.
Like anything else, it depends on what you're working on. For small personal projects I agree, but in maintaining large scale business applications with millions of lines of code, I am still having to guide the models to reuse existing code and to be more flexible with their data structures to accommodate changing business requirements in the future.
They only see a snapshot of the world the application exists in due to their ephemeral nature, and until that is overcome, they will only be able to learn general principles and not the particular nuances of your codebase that only arise from observing how the users use it on a daily basis.
That said, I think we need to be prepared for how a simple prompt like "make it faster" will work in a relatively short timeframe - and where the AI is able to adequately refactor/re-platform/be done in a way where even the architecture becomes obscure to us.
Perhaps I could set up a workflow to handle this without reading the code. I’ve seen people who make Claude fill out arcane templates, and then spend their time reviewing reports and tinkering with loops or whatever. But why would I want to read a TPS report about the code rather than just reading the code?
AI build me a cost overview of xyz: works perfectly without ever looking at the code
AI build me unreal engine 5 level graphics, or a database, or infra at high scale: this will need some active steering for now :)
Wow, what a simplistic, uncalled for view of the world. This is just so ignorantly and offensively incorrect.
This in particular is a really silly hill to die on, because literally every debate about software development, from AI to language choice to tabs vs. spaces, eventually settles into a gravity well of asserting that all the good developers agree with your take.
Stating that people can't love software development or solving problems because they are resistant to AI taking over their entire job is a silly hill to die on?
I really enjoy what I do, and have done, for the past 20 years. When I first wrote code I literally thought I discovered magic. In many ways, AI has re-triggered some of those feelings!
I also really care about building quality software for our customers.
And it infuriates me to no end when I see huge amounts of objectively incorrect code being written by AI, and the developer behind it stating they have no idea what it's doing, or when people say they can't solve a problem because they are out of AI tokens, etc.
But supposedly these are the people who care about solving problems? Not just that - they are the only people who care about solving problems.
---
Yeah, it's offensive.