Will Claude Code ruin our team?
justinjackson.caI don't know why people are talking so theoretically. This was months ago.
My friends have startups, I know a lot of engineers. The startups have been laying off people for months, and many of my engineer friends don't have jobs anymore.
Teams are already ruined. I just don't think the companies are. In many cases this seems like rational reallocation of capital to AI, and in a VC funded ecosystem you're failing at your job if you're not following the math.
I think you must have a very cushy job if you're still armchair speculating about this.
If a startup is laying off engineers then it’s dead in the water. That means it’s not growing and focused on cost cutting at the expense of velocity. Thats what a large company does. The issue isn’t AI but the startup fundamentally being broken and this being a last gasp for air before it dies.
Yeah what a lot of people are missing here is tons of small startups are laying people off, but it's not because they don't need engineers, it's because they are out of runway because their entire vertical (usually some sort of SaaS, often b2b SaaS) is basically now nonexistent. Traditionally businesses favored buying software over building it for cost reasons. Now they can cheaply build exactly what they want instead of paying through the teeth for something that is only slightly like what they want. This doesn't mean the work is gone, but it does largely mean large swathes of the SaaS vertical will be gone. The work itself is shifting to the individual businesses that were once the customers of the SaaS.
SWEs will be fine, all these small VC-funded startups building another CRUD app will not.
I work for a startup that makes a b2b SaaS that is _way_ too complex for anyone to spec out in a markdown file, especially when taking things like ITAR compliance into consideration.
We have seen steady growth and there’s been no signs of slowing down.
Our software facilitates order/quote/factory floor workflow automation with auditable trails in the manufacturing space, with cad file analysis and complex procedural pricing equations for quote generation, alongside a Shopify style storefront and many more goodies. We interface with things like shipping, taxes, erp integrations, and so much more.
I don’t see anyone vibe coding an alternative to our software even if they could. Manufacturers have enough on their plate managing their factory floors.
That said, we facilitate $millions in manufacturing orders per week and our engineering team is 3 people. We couldn’t do what we do without AI, and we would have needed to hire more engineers to handle the scale of our business if it weren’t for the power of Claude Code and Cursor.
Or there's just not enough parallelizable work for the business model...
Which again means lack of growth.
A poster-child startup is one that has a long waitlist of willing future customers, and whose engineering team is scaling the tech up, up, up to keep up with the demand.
We have yet to hit this phase in the cycle: "Hey we laid off all our engineers 6 months ago and vibe-coded this thing and now it's super buggy and AI can't fix it. Can you (senior engineer consultant) look under the hood and fix it?"
Senior engineer looks under the hood, sees 500k lines of incomprehensible spaghetti mess with emoji comments everywhere, runs out the door and never looks back.
> Senior engineer looks under the hood, sees 500k lines of incomprehensible spaghetti mess with emoji comments everywhere, runs out the door and never looks back.
Senior engineering _consultant_ looks at those 500k lines of incomprehensible spaghetti mess and sees $$$: months or years of contracts and likely very dysfunctional management that is willing to pay multiple times the cost of full time employees to keep the burn on a non-payroll line and/or keep the “AI first” story rolling on.
> Senior engineering _consultant_ looks at those 500k lines of incomprehensible spaghetti mess and sees $$$: months or years of contracts and likely very dysfunctional management that is willing to pay multiple times the cost of full time employees to keep the burn on a non-payroll line and/or keep the “AI first” story rolling on.
That's not been my experience. Even pre-AI, when I was asked to find a bug in some hacked-together codebase, sticker shock was often the result.
"What do you mean, billing for a week? The guy who created this is an actual software engineer and you're billing just as much as he did!"
I've got a list of small ex-clients who won't get work from me anymore, unless they are happy with "Here's my weekly rate, 1 week minimum".
Hourly rates don't work on a client who considers $200/m to be overpaying for s/ware development services.
If that’s how this works out then perhaps Accenture will be okay after all.
I clean up vibe code as a senior engineer consultant, it's quite lucrative actually. I specifically offer this service because I know how to do it.
I've been thinking of jumping into this sooner rather than later because I see this becoming a Thing eventually. Are you enjoying it?
It's fine, I suppose. It's like a puzzle, and you really need to be comfortable with banging your head against a wall trying to make work what is essentially immediately created legacy code by the LLM.
Works it (edit sp: Would it) help if the coders' prompts were available (checked in)?
If code is your documentation I assume it is hard to divine intent?
Prompts wouldn't really help, no. You'll have a high level description based on what you're being hired for then the code tells the story.
Well in 2 years AI will have so advanced that none of this matters.
Are you sure that AI companies will keep the current progress?
Aaaannnnd they're out of business and it was because of slowing demand and tightening credit the whole time.
Here in Europe this is not a thing, I've been hearing about such cases mostly from the US where it's clear that there is a recession going - I don't know why this is not blatantly obvious to everyone who does not view reality as whatever is said by the talking heads on TV.
I think it's less armchair speculating about the observable outcome, that people are losing their jobs, but the why. AI coding tools aren't making 10x developers, they aren't even making 1.5x developers. They also aren't making "PMs who code" or "designers who code."
It would be really cool if this was the case, I would be singing the praises of these tools finally realizing Stallman's dream of end users who can take control of all the software in their lives for their own benefit. And the huge gains we would see in open source where "man I wish there was a tool that could…" becomes "I'm gonna make a tool that…"
So personally I think it's just a continuation of the belt tightening that was and still is occurring across the economy. I don't think our industry is particularly special on this, everyone is trying to cut headcount right now.
> they aren't even making 1.5x developers
I won't try to speak for anyone other than myself, but my multiplier is definitely over 1.5x, probably higher than 5x.
I choose to sit on my hands in my freed up time so upper management does not catch on to and exploit this fact. Eventually they will though via overzealous coworkers.
> I won't try to speak for anyone other than myself, but my multiplier is definitely over 1.5x, probably higher than 5x.
So you now complete in a single Monday what used to take you Monday-Friday?
Can you even review that fast? How many LoC per day are you generating?
More like a day's worth of coding condensed into half an hour. Time to review/test is mostly unchanged.
Day to day is mainly minor feature additions into a stable product so not a huge amount of code churn.
It’s easy to produce a high volume of code, sure, but it is not equally easy to test, verify, and integrate it. And with a high volume of code, there is a high volume of shit to review & test & integrate. For companies that give a shit about not vibe coding their way into a disaster (because they have lucrative enterprise contracts that depend on reliability & security), that’s the real blocker. (Plus, these types of projects are big, not trivial, and things are harder to integrate & properly test because of that.)
Not to mention, if a team wants to keep a semblance of understanding of what they own & ship… it can be exhausting to have a huge volume of new code coming into the system.
It’s definitely a productivity unlock. For sure. But there are a lot of knock-on effects we’re still figuring out that counteract how much extra “value” we’re shipping
In my case, the volume of code is roughly the same. I'm not using the efficiency towards pumping out more code, just using it to be AFK more.
I spend enough time iterating and refining to the point I'm comfortable taking ownership of the outputted code. Perhaps hypocritically, I do mald when people upload code for review that they clearly haven't taken the effort to read through critically.
It’s not hard to 5x your productivity when you were close to zero to start with.
Is that you, boss?
People with a lower multiplier are either in the minority of developers solving genuinely hard/novel problems or, more likely, they've just not figured out how to tap into AI's potential.
Granted, to your point, a decent chunk of the HN crowd belongs to the former and can't relate to us paycheck stealers.
I always hear people say this, but it’s not clear to me what exactly is so difficult about using AI that otherwise-competent developers “can’t figure it out”
My hunch is it's a combination of
* coming in with a bias of not wanting it to work
* having too high of an expectation
* giving up too early
* not trying SOTA models
* not taking the effort to communicate intuitive or painfully obvious things
But perhaps it is too dumb to solve the type of problems you guys are working on and no amount of cajoling will help. All I know is "it works for me."
Agreed, the tech job market was bad before AI was useful.
The "I'm gonna make a tool" thing is slowly happening and will probably help Linux, knocking on wood... https://x.com/xpasky/status/2030016470730658181
As an engineer and a daily if not hourly user of Claude Code, I would never dream I could do the jobs of my product / designer teammates. Not because I don’t have opinions on product or design, but simply because they do me the huge service of attending meetings so I don’t have to.
I recognize the necessary evil that is Zoom calls and face-to-face time in the larger context of running a business, but I also know what I’m good at and what I’m not. And long, drawn-out “alignment sessions” are not in my wheelhouse. If my PM and design friends are happy to take that bullet for me, I’m happy to let them do so.
When the product marketer, product manager, and designer are the same person there is no alignment meeting.
In my experience at BigCorp, Inc., "alignment meetings" are for meeting with other teams, not for meeting with your own team. You'd just call that a "team meeting", "feature crew meeting", "leads meeting" (if it's just leads), or something to that effect. In BigCorp, Inc. you traditionally need to have both kinds of meetings somewhat regularly. Not sure how the advent of AI will change that structure.
They meet with customers as often as they meet with each other. Another one of those necessary evils IMO- it's not fun but it needs to get done.
The final frontier: a warm body.
I think the reality is that these tools are good enough that, to some degree, all three of the roles are correct, everyone is now definitely more able to do others' roles, leadership knows this, and even if there's a period of inefficiency or overwork or lower quality output, there's going to be a drive toward collapsing responsibilities. OpenAI saw this early: Member of Technical Staff. The degree to which this negatively impacts the team or company's output is really a function of (1) how drastically leadership does layoffs, (2) how quickly models and agents continue to improve, and (3) how earnestly leadership can admit mistakes and backfill humans when they realize they've over-fired.
In other words: Yes it will ruin our team.
I won’t take credit for this insight, but as someone else pointed out, everyone oversimplifies other people’s jobs. To PMs engineers are just code monkeys that they won’t need soon. To engineers, PMs are the guys that manage Jira. Designers are the fussy people that make things look pretty. The reality is all these jobs have intricacies AI absolutely sucks at but those intricacies are lost in the larger discussion.
As a coder though, I’ll point out this is why the “AI solved coding” shit drives me crazy. You only believe that if you don’t know how to code or you have an agenda.
Instead of engineer vs PM/manager I separate those jobs based on these categories.
1. How long they can survive in the job while being mediocre or outright bad at their job.
2. Probability of failing upwards.
Engineering roles tend to filter out bad candidates more early, quickly and the probability of failing upwards is less when compared to PM and managerial roles.
Also, in my experience PM and managerial roles looks like skills based jobs but they tend to select individuals with specific personality types and they are more likely to excel.
Developer roles also select towards certain personality types but I think its more diverse than we care to admit.
And for people who never liked it in the first place and just wanted a quick buck.
You can't pay Claude with equity though
You can tell him and he will love it and I bet he has encoded the behavior of working better because of it just how some tried threatening LLMs to work harder
which doesn't make sense because the quick buck train is gonna stop. But then again, maybe there wasn't any sense in them in the first place.
No. There are intricacies to every job but these are intricacies that are learnable. It’s like learning how to drive. It’s a skill for sure but anyone can do it after practice.
I thought programming was the same thing for a long time but have grown to find out that this is not the case. There are many people who cannot learn programming in a reasonable amount of time and therefore are unable to pick up the skill. It is not universal like car driving.
The thing with being a PM or a designer is that this skill is learnable. Anyone can do it. The reason why these jobs are segregated is because society is under the delusion that these are special skills that require intense training when at most he training is equivalent to learning how to drive.
Some of you may be thinking I’m insane but there are tons of jobs that are like this. The presidency for example. You can be senile and insane and still be president. The country doesn’t blow up just because you’re insane. Or maybe this isn’t a good characterization.
Hmm electrician or plumber is the better comparison. The skill level required to be a PM or designer is equivalent to electrician or plumber. Anyone can pick it up with training. It’s not rocket science folks.
Huh as someone who has a career doing both design and engineering I disagree with this take.
I think both skills can be learned. I also think that people have intrinsic talents that make them better or worse at those skills.
Put another way, anyone can learn to code but some people will never be great at it while others have a natural talent. Same for design.
I’m curious why you think otherwise. What’s the difference in your mind?
I’m kinda getting off topic here but anecdotally: I’ve tried to get so many of my friends to learn programming. I love it, and I think a lot of em would love it too. But they hit a hard wall with the patience needed to self learn.
Like the moment something doesn’t happen like the tutorial said (error message saying “idk what python is, you mean python3?”), they just give up completely instead of googling it. I really feel like the venn diagram of “people who can code” and “people who can google errors they don’t understand for a couple hours” is nearly a perfect circle.
LLMs can smooth out those little tediums, so maybe more people really will be able to learn programming now. But then again if you don’t have the patience to trudge through the annoying parts, will you have the patience to be confused and struggle, instead of letting Claude do the hard stuff for you? I’ll be interested to see what future self-taught devs look like!
Your friends struggle with learning programming because they don't care enough about learning it. You're the only one that cares.
Same can be said for any skill.
Threads like this bother me a bit because it makes programmers seem so smug, like they are this gifted class that is able to wizard the machine where mere mortals cannot.
Its intellectual elitism.
Regardless of that. I do think it's true that not anyone can learn it.
But the "elitism" is becoming something that is less and less relevant because people are less needing to learn it anymore thanks AI.
What I have observed is, if you don't know what the issue is, llm would usually suggest something that is unnecessarily complex and not ideal.
It might work but the moment something fails, llm suggest hacks instead of solution.
Hard agree here. I think the best predictor of whether someone will be good, eventually, at something is “do they love it”. If they do then chances are they will spend lots of focused time practicing and actively seeking out ways to get better.
Maybe that love, or at least liking something, comes from inherent talent to some degree but all the talent in the world won’t help you if you don’t put in the time.
Maybe you can learn to be mediocre or good enough designer or similarly good enough or mediocre engineer. But I don't think you can actually learn to be a great designer or great engineer - it just takes a different set of skills and evolutionary and genetic material which isn't available to all the people. Some people are simply not good enough in logical and abstract reasoning, math, and similar range of skills which are essential for becoming a good engineer. Similarly someone who is a good engineer doesn't and likely can't have skills required to become a good designer, it just takes a brain which is hardwired and developed differently.
As someone who considers themselves a good engineer and a good designer (and does both professionally) I hope that last bit isn’t true
Sure, this is why I said likely. If you are good in both then I believe this isn't something commonly found in people.
Hard disagree. Design is easy and trivial. Thinking it's hard is a sort of mass delusion. Engineering is the hard thing.
Easy. I met people who tried really hard to learn how to code and failed.
Design on the other hand especially modern design is easy. It's just text placement, geometric shapes and proper colors that synergize. This isn't like anatomical drawings or oil paintings. It's not just easy, it's obviously easy. What needs to be learned is how to use the tools and do it with speed which does take time and training, but again this is not rocket science, a lot of what looks "good" and "modern" is intuitive and obvious. And modern design is just easy to draw.
I mean look at hacker news. It's pretty clean. I like the aesthetic. I bet a "designer" didn't even touch it.
This shows a deep misunderstanding of what design is.
What you’ve described is “visual design” which is a subset of the design field.
There are many sub-specialties, but at its core design is about problem-solving, communication, and empathy.
There are a lot of bad designers who are great at making things pretty.
A good designer spends more time researching, understanding the problem space, interviewing users, brainstorming, etc., than pushing pixels around.
The context of this thread is visual design for websites. Not design for off topic bullshit like furniture.
Also I agree with a lot of what you said. The only difference is I feel anyone can do it. The qualities you attributed to a good designer are trivial to learn. Make no mistake it takes time and effort to do these things and many companies neeed a specialized role where someone is only doing this thing…
But anyone can do it and learn it. And not anyone can learn how to program.
> It’s like learning how to drive. It’s a skill for sure but anyone can do it after practice.
The analogy only illustrates the parent's point. Most licensed drivers have been doing it for years and are still terrible drivers, because they never grasp the intricacies of driving — smoothly accelerating and decelerating, smoothing out corners, anticipating light changes, gauging merge distances and timings, using mirrors well, ensuring cars get by when making a left turn in an intersection, etc, etc
>The analogy only illustrates the parent's point. Most licensed drivers have been doing it for years and are still terrible drivers, because they never grasp the intricacies of driving — smoothly accelerating and decelerating, smoothing out corners, anticipating light changes, gauging merge distances and timings, using mirrors well, ensuring cars get by when making a left turn in an intersection, etc, etc
This is not true at all. Most drivers pick this up. You only tend to see this with beginning drivers and it eventually becomes better. The overwhelming majority of people learn how to drive and they learn how to drive quite well.
>designer is that this skill is learnable
Why do you think this? Being a designer is ultimately a matter of "good taste" and intuition for HIG (that you learn to systematize and formalize) and not everyone has this to start off with. Lack of good taste is how you get stuff like liquid glass. People can learn to compensate for lack of good intuition, but it's the same as someone without innate mathematical aptitude compensating for intuition by "grinding through the algebra".
>This is an absurd take. Everyone looks at the other side and says, "Yeah I could do that". Few can.
Clearly Apple pays someone tons of money and calls them a designer and pays you nothing because your anti liquid glass opinion is shit to them. Coming here and talking as if your opinion on what "good design" is the end all be all lacks nuance and perspective.
Overall design at it's core is text placement along with media like pictures or graphs. And then putting it along with interactive buttons. It's a form of art, and this art is significantly easier than say figure drawing or painting with realism. The main reason is because the shapes used are simple, no complex shading, no need to really think about how light shines on a complex surface or how clothe interacts with the human body. Just some flat colors some typography, placement of some buttons to transition to other pages and that's it.
People are biased, but all you need to do is stare at the mona lisa, and then at the spotify logo and it's totally obvious that one took skill and the other took just just some thought. Design is easy. Liquid glass ironically is the one modern design trope that would actually take skill to render by hand, but of course we have algorithms doing it so it's more the skill of the programmer than an artist.
> I thought programming was the same thing for a long time but have grown to find out that this is not the case.
> The thing with being a PM or a designer is that this skill is learnable.
This is an absurd take. Everyone looks at the other side and says, "Yeah I could do that". Few can.
>This is an absurd take. Everyone looks at the other side and says, "Yeah I could do that". Few can.
For designers, all you need to do is look at the google logo. That's the epitome of skill in design. It's trivial to come up with a multi colored word and have it fit in a sort of clean aesthetic.
For PMs it's much more harder to prove it to you. So I won't go about this useless endeavor. Suffice to say it's not really automatically "absurd". It's more likely you have some sort of identity and connection with being a PM and you're proud of that identity and your attack here is a defense.
If you truly were disagreeing with me in an unbiased way, you wouldn't just call it "absurd".
Are you mistaking your interactions with low level trades roles (the guy who's making bank fixing power sockets on the weekend) with say, the people maintaining factory electrical systems? or designing them?
Nah more like the guy who wires up the whole house.
Factory electrical systems are on another level than your typical 110V AC. But given what I know I would say that the factory electrician sits right at the borderline between comparable to SWE and anyone can learn it.
One of the more ignorant comments I've read on HN which is saying something.
No, the ignorance is with you. It's obvious. Ignorance is evident in how a person approaches an opposing opinion. Does the person question it or does the person label the other as "ignorant" and then move on? The later is actually the one that's ignorant, and that's you.
Do people really think like that? I see the other people at my company as human beings solving complex problems whether they are an engineer, a manager, an exec, or HR.
Absolutely. This is precisely why people are saying AI will eliminate software jobs.
Do we need more of those “If you think it’s so easy, try being me for a day!” sitcom episodes?
Being a generalist is less fun when you don't have specialist colleagues around to teach you new things and take over the tasks that require actual experience, training, intuition, etc.
AI in general is ruining companies as a whole.
Companies are finding out the hard way that replacing engineers with AI, is costing them twice as much to fix the problem and having to hire people again.
If I get fired because a company went crazy into AI, things went bad, and they call me back, you sure as hell I am not returning.
Hey there! This is the exact thing that happened (although I cannot say I was a perfect employee). They replaced me for a relatively important product, where I was already the only dev along with 4 PMs that never agreed on each other.
The entire company is hell, it made that switch overnight and never looked back at the devs that have to endure all this bullshit.
Luckily I found another company that treats devs as humans. I've convinced one other employee at $EX_COMPANY to make the switch, with success! I'm planning to find everyone there a proper job that doesn't make u feel like you're under surveillance.
This naturally seems to indicate smaller teams will become more normal.
Many of my clients are blown away by what our teams can do with 1 senior engineer now.
Anything below enterprise level software should be thinking very hard about what team composition actually needs to look like to achieve good results. It's likely a lot smaller headcount than it used to be.
I definitely think teams will be getting smaller, and will become even more outcome oriented.
One phrase stuck with me from this: "they'll start to absorb lessons that it took their colleagues decades to learn."
I think the point of failure now will hinge on the willingness of teams to admit what they don't know. The ones that don't won't be saved by Claude.
What’s changed now, with AI, is that you can have compressed learning cycles. Folks can build, deploy, and learn faster now than ever before.
Admitting when you don’t know something has always been important; but the ability to build, deploy, and find out has never been greater.
Instead of theorizing about what might work, you can just build it and find out.
> Folks can build, deploy, and learn faster now than ever before.
Certainly build and deploy faster. You aren't going to learn faster though.
Just like reading math without doing problems doesn't enable to you pass your exam, reading code without writing any of it doesn't allow you to learn at all.
> Folks can build, deploy, and learn faster now than ever before
Fail faster yes. Learn faster no. The research out there shows that having the AI doing the work stops the learning process.
I use the GenAI tools all the time and I'll be the first to admit that cognitive debt is a real thing
> absorb lessons
That maybe correct for some lessons. Many lessons you have to learn the hard way to really absorb them.
I wonder if this could go down similar way some SaaS systems went to lower barrier to customisation so less technical users could do it. For example, having to interface with some ServiceNow instances I often find major flaws with db schema design - similar data in multiple places, lack of constraints, etc. Basically one big mess you are now stuck with that could have been avoided if a db expert was in charge of data model design.
I’m mainly concerned about communication worsening. If a former engineer, designer and PM all want to do the same job now, the results will be fine, excellent, even, but only if they communicate and build together better than they used to. People who only communicated along the lines provided by the design of their role or organization have new skills to learn.
sounds fun if framed as: everyone now has tools to help each other's role more
XP is still a good method to practice, but now with AI.
The class system of business is a warning here.
There used to be hundreds of humans doing math by hand. They were computers. The people that managed those armies of humans were management class.
Then came actual silicon computers. The ones that managed those, despite the fact the value, quality, efficiency, productivity of the systems they now managed dwarfed the old human armies, those people were no longer management. They were labor.
AI will bring a similar effect. These front line "managers" who were already greyarea management, will be labelled "labor".
The skills list weirdly leaves out “knowing the capabilities and limitations of the tools.” This is non-trivial to say the least.
The roles list also leaves out testing, which seems to me to be the second most important thing (after specifying). This may be because non-testers assume that testing is easy or will be done by the AI. But any testing done by AI is not testing at all, because the effect of real testing is to inform a human of the status of the product based on that human’s empirical investigation. When AI “tests” the humans are being asked to trust instead of investigate.
Three parallel terminals of coding agents and crosswords magazine until they complete their jobs. Work with real experts and solve real world problems.
I think it will enhance everyone.
PMs can't develop, since llm development (adding code to whatever the llm initially spat out) still consumes time and effort, but they can now write a PoC without devs and quickly get it up and running without sys ops.
It already has. Those jobs PM, designer and engineer are all affected and there will not be a need for too many of them for a product feature and will be reduced over time.
While those roles will still exist, there will be a initial shock in people who once believed they were 'valuable' but the business thinks otherwise and does mass layoffs just like Block, because of let's face it; AI.
The way to still remain relevant is to absorb all three roles and build a startup with Claude Code on your side and move rapidly.
> The way to still remain relevant is to absorb all three roles and build a startup with Claude Code on your side and move rapidly.
Just how fast can you possibly move? I can take a gander at your good idea, get claude to implement it in less time than it took you, and offer it cheaper as a result.
I think you're probably still going to need all of those people, just less of them to accomplish the same amount of work. It's the nature of LLMs that they will always need oversight. Having 20 or even 100 of them agree on something is never going to be 100% accurate -- it can't, the randomness and hallucinations are part of the secret sauce. There will always be a need for people to verify what they do. That said, it was the exact same when people were doing the work.
You need to be sociopathic aggressive right now. We are in a few month window here where execs are very confused even if they parrot their group chat talking points to sound smart and ahead of it. They are all behind and they know it.
This isn’t comfortable but now is the time to ship fast and hard. To overstep boundaries and be the person getting attention. In a few months everyone will be so you need to do this now.
Just don’t. Don’t limit yourself. Ask for forgiveness.