The "just build it with Claude" paradox
18 comments
Totally agree and basically is destroying the software economy that we have, because it feels that everyone can make their own app, suiting their own requirements. But they don't realise this to be a trap, because the majority of the effort of any product is the ongoing maintenance and evolution. Let it a new browser version come along, a new OS version, a new LLM version and once their "tuned", self-made app suddenly stops working or misbehaving, they will realize that it ain't so easy as promised, even if the LLM can be used again to evolve it.
But as all major breakthroughs, the path is forward and there is no logic argument that you can make to let people consider otherwise. Eventually, all the dust will settle down and it will be easier to uncover this and other misconceptions, until then, no worth trying to convince people otherwise.
But as all major breakthroughs, the path is forward and there is no logic argument that you can make to let people consider otherwise. Eventually, all the dust will settle down and it will be easier to uncover this and other misconceptions, until then, no worth trying to convince people otherwise.
I mean there's a world where generative ai becomes good enough to literally do it all, and then we're all screwed. But let's assume that's not the case as it isn't today. It feels like there's no point of reflection. Do people just continue to value their time less and less? Does the problems we sell need to be THAT difficult? It feels like in a few years the only problems worth solving will be problems where we need phd level engineers to solve.
I have noticed a similar thing, but from an IC culture like (not direct quotes, but the sentiments align)
1. Manager: "This should be fast and easy, just Claude it, why is it taking so long, didn't you use claude?"
2. Teammates: "Dont ask me questions, ask Claude. You should be able to teach and onboard yourself"
3. Code Reviewer: "I don't really know the codebase, but claude says this doesnt work so you must fix to agree with claude before I will stamp"
1. Manager: "This should be fast and easy, just Claude it, why is it taking so long, didn't you use claude?"
2. Teammates: "Dont ask me questions, ask Claude. You should be able to teach and onboard yourself"
3. Code Reviewer: "I don't really know the codebase, but claude says this doesnt work so you must fix to agree with claude before I will stamp"
> Lets say on the low side for their specific use case they spend ~10 hours. That’s $1k of their time
Only if there is perfectly elastic demand for their time at the given rate. It's often the case that for people putting in a typical 35–45 hours Monday through Friday, the number of hours in a week is not the limiting reagent, though (else they would be putting in 50, 60, or 70+ hours instead).
Only if there is perfectly elastic demand for their time at the given rate. It's often the case that for people putting in a typical 35–45 hours Monday through Friday, the number of hours in a week is not the limiting reagent, though (else they would be putting in 50, 60, or 70+ hours instead).
> The thing is, the pain I’m solving is real. People just want to solve it themselves.
It is not that painful if they can solve it themselves though and are not immediately interested in a solution.
The other problem is that for most people, they are doing Claude building in parallel to other work. It taking 10 hours of Claude prompting is not the same as it taking 10 hours of my time.
It is not that painful if they can solve it themselves though and are not immediately interested in a solution.
The other problem is that for most people, they are doing Claude building in parallel to other work. It taking 10 hours of Claude prompting is not the same as it taking 10 hours of my time.
What if they can't actually solve it themselves though? The solution I'm building works significantly better than something you can throw together with n8n in a few hours, but people's perceptions are all that matters. The pain is real, but the way we measure success (especially for consumers) doesn't necessarily have to match reality.
I should also state, I have customers and I have competitors. This isn't necessarily something that isn't worth solving. I'm just noticing over the last 6 months it's becoming increasingly common for people to 'believe' they can do this with claude more and more. Whether they can or not, and whether its worth their time or not, the perception is increasingly that the value of one's own time doesn't matter.
I should also state, I have customers and I have competitors. This isn't necessarily something that isn't worth solving. I'm just noticing over the last 6 months it's becoming increasingly common for people to 'believe' they can do this with claude more and more. Whether they can or not, and whether its worth their time or not, the perception is increasingly that the value of one's own time doesn't matter.
Then they presumably either crawl back to you or the pain is in fact not real.
Getting familiar with AI and even building with it is valuable in itself. Every beginning of a new process takes time but eventually you will build systems that will save you time.
Sure but does this mean that consumer saas or anything that's more trivial then thousands of hours of eng work is essentially dead?
What is "consumer SaaS" anyway? Do you mean B2C with "consumer"? Because no, B2C is not at more risk than B2B.
Consumer saas being anything like a real consumer product. Games, fintech apps, task apps, etc. My product specifically is in a grey area of b2b vs b2c because it's for early stage founders/pre revenue companies. So arguably b2b but very upstream.
Yeah, you're in the worst hit category by far. Both for B2C and enterprise B2B there are reasons that mitigate it a lot.
Eh, I haven’t found that to be the case. The AI workflows are too nascent, too stochastic, and change too often to be a good use of time.
Subscription fatigue is real, especially since most SAAS providers got a lot more aggressive with raising prices in the past few years.
Add enshittification, products randomly getting acquired and shut down, etc, and $50/month quickly stops looking like $50/month.
Not saying the math works out for the “build your own” version… But it’s not that clean for the subscription option either.
On a separate note, there is the concept of “ideal customer profile”. If the people you’re speaking with feel they can solve this problem on their own, then they are not it!
Add enshittification, products randomly getting acquired and shut down, etc, and $50/month quickly stops looking like $50/month.
Not saying the math works out for the “build your own” version… But it’s not that clean for the subscription option either.
On a separate note, there is the concept of “ideal customer profile”. If the people you’re speaking with feel they can solve this problem on their own, then they are not it!
This is kind of a really interesting thing I haven't thought about... My ICP is founders, people using lovable/etc, and people trying to get their first dollars. It's a marketing and distribution product, which a lot of founders struggle doing, but you raise a really good point that my ICP is literally the same person that thinks they can build everything themselves.
Yeah, that is a tough market… By definition those people likely have very little cash, and higher than average belief in their own abilities :)
I keep running into “why can’t I just build this with Claude?”
And it’s like… yeah, you probably can? I’ve put ~100 hours into getting an MVP working and trying to market it at around 30-50 bucks a month. Then I’ll talk to someone making 100 an hour and they go, “yeah this is a good idea but I’ll just hack this together in a few weeks with Claude and n8n”
Lets say on the low side for their specific use case they spend ~10 hours. That’s $1k of their time to avoid paying 30 bucks a month.
On paper that makes zero sense. But it keeps happening. The Claude sub is already paid for, so the output feels free. Meanwhile paying for a product, even a cheap one, feels like a real decision.
So you get this weird outcome where people will happily spend way more in time to avoid spending a tiny bit of money.
Everyone keeps telling me to focus on real pain points. The thing is, the pain I’m solving is real. People just want to solve it themselves. That always made sense in enterprise b2b space when you have an engineering team and there are more tradeoffs than just time vs money for build vs buy. But now I’m seeing founders and consumers do the same thing, and I’ve never seen people treat their own time as this expendable.
Not sure what to do with that yet. Curious if others are seeing the same thing.