AI is causing everything to change, small basically trivial oss projects are no longer really needed - that doesn’t seem too bad, we can focus on the more exciting things now!
I’ve often wondered the opposite. The number of pieces of software I used to pay for that I was able to replace with software I’ve built using agents is nearing double digits. I wonder how much longer people will need to pay huge software companies instead of just collaborate with a few other folks using AI and build their own better version.
As a developer (it sounds like), you were probably never even a drop in the bucket dumped into the ocean of the proprietary software market. Most people don’t want to change their own oil.
Not only you're making huge assumptions on AI's ability to build software useable by non-software engineers without professional input, you're also underestimating people's desire to pay just for convenience ("gee, I need a diet tracker, let me find a few people to help me vibe code it instead of paying $10")
It's a valid assumption, and it is going to have a huge impact on the SaaS marketplace.
"I need a resource tracking application. Let me find a consultant to help me vibe code it rather than pay $10,000 a month."
I have paid a relatively big price for a software which does a simple thing: Drawing diagrams. The application? It's called OmniGraffle pro. That simple looking program drawing diagrams can do flowcharts to scaled architectural or industrial diagrams and everything in between.
Try replicating it with a couple of agents. Not the capabilities, but the performance, stability and dependability and sustainability as well.
Same is true for my favorite text editor on macOS, BBEdit, or IDE of choice, Eclipse. I can increase the number of examples.
These tools are valuable not because they are expensive. They are valuable, or indeed priceless, because of the polish they have. The small features make no sense for a beginner or how they cope when things get absurdly haywire.
Immediate features make us excited and they are easier to replicate because they are visible, but things like performance, dependability, error recovery or the small tweaks they have gone through in the last decade are not, and this is what matters.
We're exploring computing via AI again, and we act that we did nothing in the last 60 years or so. Software and computing is much deeper than it looks at the surface, and we'll learn this in a very very painful way, again.
As I've been vibe coding a side project, I've been curious on just how much cruft and slop the agents are training on now. A part of me feels like I should actually have it be in a private repo.
I've noticed that for smaller projects (<=10k lines of code) the valuable thing is no longer source or binary artifacts, it's the encoded ideas, architecture and domain knowledge. A folder of plan files is about as valuable as the code itself, and easier to adapt to new domains
I wonder if that will spawn a new type of open source project. Where we have free exchange of ideas, without necessarily requiring the free exchange of code (which would violate a lot of Stallmann's ideas about open source, but might ironically be attractive to people who dislike AI training on their code)
For one off tools, sure. For ongoing tools, either they are right after the first implementation and you use them for years. Alternatively, you need to maintain them and they decay until you discard them.
You also need to be ok with your tool to do task A (you had an agent build) to work different from my tool to do task A (I had an agent build). Meaning we can’t reliably compare the output from those tools.
The ecosystem of open-source reusable components will remain a good idea as long as architectural principles like low coupling and high cohesion still apply—otherwise it’s wasting resources on reinventing wheels.
AI is causing everything to change, small basically trivial oss projects are no longer really needed - that doesn’t seem too bad, we can focus on the more exciting things now!
I’ve often wondered the opposite. The number of pieces of software I used to pay for that I was able to replace with software I’ve built using agents is nearing double digits. I wonder how much longer people will need to pay huge software companies instead of just collaborate with a few other folks using AI and build their own better version.
As a developer (it sounds like), you were probably never even a drop in the bucket dumped into the ocean of the proprietary software market. Most people don’t want to change their own oil.
Developers don’t even like to pay for their own code editors. Something they used to work in all day.
Roughly zero.
Not only you're making huge assumptions on AI's ability to build software useable by non-software engineers without professional input, you're also underestimating people's desire to pay just for convenience ("gee, I need a diet tracker, let me find a few people to help me vibe code it instead of paying $10")
It's a valid assumption, and it is going to have a huge impact on the SaaS marketplace. "I need a resource tracking application. Let me find a consultant to help me vibe code it rather than pay $10,000 a month."
I have paid a relatively big price for a software which does a simple thing: Drawing diagrams. The application? It's called OmniGraffle pro. That simple looking program drawing diagrams can do flowcharts to scaled architectural or industrial diagrams and everything in between.
Try replicating it with a couple of agents. Not the capabilities, but the performance, stability and dependability and sustainability as well.
Same is true for my favorite text editor on macOS, BBEdit, or IDE of choice, Eclipse. I can increase the number of examples.
These tools are valuable not because they are expensive. They are valuable, or indeed priceless, because of the polish they have. The small features make no sense for a beginner or how they cope when things get absurdly haywire.
Immediate features make us excited and they are easier to replicate because they are visible, but things like performance, dependability, error recovery or the small tweaks they have gone through in the last decade are not, and this is what matters.
We're exploring computing via AI again, and we act that we did nothing in the last 60 years or so. Software and computing is much deeper than it looks at the surface, and we'll learn this in a very very painful way, again.
As I've been vibe coding a side project, I've been curious on just how much cruft and slop the agents are training on now. A part of me feels like I should actually have it be in a private repo.
I've noticed that for smaller projects (<=10k lines of code) the valuable thing is no longer source or binary artifacts, it's the encoded ideas, architecture and domain knowledge. A folder of plan files is about as valuable as the code itself, and easier to adapt to new domains
I wonder if that will spawn a new type of open source project. Where we have free exchange of ideas, without necessarily requiring the free exchange of code (which would violate a lot of Stallmann's ideas about open source, but might ironically be attractive to people who dislike AI training on their code)
For one off tools, sure. For ongoing tools, either they are right after the first implementation and you use them for years. Alternatively, you need to maintain them and they decay until you discard them.
You also need to be ok with your tool to do task A (you had an agent build) to work different from my tool to do task A (I had an agent build). Meaning we can’t reliably compare the output from those tools.
The ecosystem of open-source reusable components will remain a good idea as long as architectural principles like low coupling and high cohesion still apply—otherwise it’s wasting resources on reinventing wheels.
[delayed]