Can AI take over XenForo development entirely future?

[Yenxji]

With the rapid advancements in AI development tools and its core of XenForo, it’s interesting to imagine what the future of software like XenForo might look like. Could we eventually reach a point where AI alone is capable of maintaining XenForo for fix bugs, releasing compatibility updates, and even building new features? So that AI can handle all without human intervention in the next 10 to 30 years away.


There is a few AI exists like GitHub Copilot and ChatGPT, VEO 3, Claude AI, Bolt New, and others that are already assisting developers with writing and debugging code. But could that AI handle future versions of these tools become advanced enough to independently take over XenForo core development possible?

 

[Kirby]

Do we really need yet another AI development thread?

XF 2.3 Thread 'AI agents'

Jun 6, 2025

Uh... it's not often that my mind is blown, but this morning I screwed around with Junie (an AI agent inside phpStorm). I gave it zero training or instruction except I just wanted to see if it could figure out how to make a XenForo entity class on it's own, so I asked it: "make entity"... literally that was it.

Somehow it decided to get into my Marketplace addon and create a coupon entity (I didn't say anything about that addon or that entity... look at step #3 here).



Details of the steps it went through:





... so literally 2 words of...

• digitalpoint
• Replies: 57
• Forum: XenForo development discussions

• 
• 

 

[Yenxji]

Do we really need yet another AI development thread?

XF 2.3 Thread 'AI agents'

Jun 6, 2025

Uh... it's not often that my mind is blown, but this morning I screwed around with Junie (an AI agent inside phpStorm). I gave it zero training or instruction except I just wanted to see if it could figure out how to make a XenForo entity class on it's own, so I asked it: "make entity"... literally that was it.

Somehow it decided to get into my Marketplace addon and create a coupon entity (I didn't say anything about that addon or that entity... look at step #3 here).



Details of the steps it went through:





... so literally 2 words of...

• digitalpoint
• Replies: 57
• Forum: XenForo development discussions

• 
• 

This is not about future of XenForo core how AI could handle and take XF core over human, this topic is talking about the making of addon development.

 

[digitalpoint]

Still a long way off before that (where it's autonomous). Would need to be to a level where AI could build XenForo from the ground up from scratch on it's own. Not happening anytime soon.

For now, it's more like a junior-level programmer that you can give boring/repetitive tasks. But just like a jr programmer, you'd want to review what it's doing.

 

[Jeremy P]

Yeah, it's difficult to be measured about. On one hand, it still requires a lot of hand-holding to do anything complicated unless you're working on a greenfield project. On the other, it's an invaluable tool for the right tasks. I think the "copilot" terminology is apt at present. It's still obvious when people try to use it to do things without a good understanding of the fundamentals. And it's impossible to speculate about how it will or won't progress with much certainty.

 

[Sim]

 

[Stuart Wright]

I have a coding background (since the 80s) and I coded for vBulletin, but XF development is too much of a steep learning curve for me. I wish there was either a training course that hand held me through learning XF addon development (I’d pay £000s for it) or an AI tool that would do it all.

 

[ebikes.com]

I run a web development shop with all senior developers and we work on hard stuff that hasn't been built before.
AI is great for reproducing things that have been done before, so we don't get good results.
We are still disappointed with what's state of the art.

AI slop can produce massive amounts of technical debt and make the code progressively harder to maintain - we only use AI strategically.
The most critical design decisions and work still require a strong systems thinker and probably will always.. unless you can figure out how to get an AI to experience everything a human does... plus make an overgrown calculator think within 99% of how a human thinks.

Until then we use AI for:

• research ( powerful first pass )
• snippets
• fast prototypes to prove a concept
• some low to mid complexity data processing tasks
• write some testing code
• code which fits in the category of: too much time needs to be invested to make it happen, and it happens to not be a critical feature, so sometimes slop is ok

Vibe coding? hell no, you are leaving a mess for future generations of programmers to rewrite from scratch, which will drag on your business