This is jut nuts https://www.youtube.com/watch?v=eFB79TYI-Vw , I sam starting to believe that software development is dead guys… s*it!
Yes, writing software manually will shrink dramatically.
Right now, one good dev + AI does what a small/medium team used to do. The value is you can understand a real-world problem, describe and prompt it precisely, test what AI produced, detect and trim bullshit and market the result.
But in the upcoming years, I am not sure if this also will even be a thing any more… we are probably watching the end of programming as the central activity of software development, maybe even more, maybe in the future you just describe what you want and the AI OS will build a tool for it on the fly and cache the result in case you will need it again.
I also think coding languages will converge to a single one, which will be optimized for agents.
We each need to figure it out how we fit in this new world…
2 Likes
I agree. The thing is that if it can do this to software imagine the rest, AI is comming for us humans, what is the advantage to use to have such a tool if it destroys all jobs, nothing will have value anynmore, a few years from now we will have robotos everywhere as well, wtf is happening to the world it feels like a SF movie!
2 Likes
This is what I’m thinking about non-stop as well, what survives this?
-> Things that are simply offline code, like Photoshop, like a plugin that adds a feature, PDF Reader, ZIP Extractor, Disk Space Analyzer, that’s probably dead
-> More complex things like Anydesk or Loom are not gonna do well either. I think I pay $180/year on Loom. Replacing it is now a 1-day project where all I need is Claude and a VPS to host a video player on. These things will not die and they have some network effects, but competition will eat their margins.
-> Ahrefs will do well, at least in the next 5-10 years, because they’re not actually selling code, they’re selling data,
-> Cloudflare might do well, because they’re selling infrastructure, not code.
-> Probably Shopify does well too, a lot of it is infrastructure,
I think the conclusion I’m reaching is that we can’t sell code anymore, we have to sell data, or infra, or something else.
I think the difference between infrastructure and code is that you’re not paying for the feature, you’re paying for the reliability.
It feels like you can no longer market yourself as “we have all these features!”, features are becoming worthless. In fact, why do i need to buy your bundle of 200 features, when I can just vibe code the 7 ones I need, and specifically to my needs, not your vision.
The value proposition has to be outside the features somehow…
Exactly!! And data may become much more expensive than infra, because AI is insanely data-hungry, but not just for more data, but for better quality and domain-specific data which is not AI generated ‘junk in - junk out’.
Check micro1 for example: they are not really selling software, they are building the platform for human evaluation and provide data to train AI models.
I think this is also a direction which might be profitable to insist on.
1 Like