I sure miss the old Xenforo days

[dvduval]

Just for your education..

Laravel, out of the box, takes about 30ms of time to simply do it's internal gymnastics routine before outputting hello world.
Xenforo, on the same CPU, can boot and have a page drawn in 2-3ms. Most of that time is spent connecting to a database and grabbing data. Otherwise it's damn lean.

Here is what Laravel performance looks like versus using a minimalist PHP framework. Xenforo's may be a few milligrams heavier than Zerolith, so it's a decent comparison point.

View attachment 341001

So just by switching your backend technology, out from Xenforo's you're going 10x slower for a simple request. That's before you do any real work.

How did that work for flarum? see their support form...
https://discuss.flarum.org/?q=performance

Don't even get me started on React. And Tailwind is a very bad idea for bandwidth reasons for forum software.

Use this combination of technologies and you will end up with a forum that is not feasible to run on the public internet at scale unless it happens to be funded by a fortune 500 company.

Fair enough. It's good to see the discussion taking place what would be the best methods to refactor/modernize xenforo. The key point here is with enough consultation with a frontier AI model, large improvement could be made quickly. Failing to start using AI bring xenforo into the modern era I believe is a big problem.

 

[duderuud]

Kier is still active daily. When I see posts flying about, I did start to doubt myself, but I believe things will continue to grow at XenForo.


Message from Kier on the 3rd June.
"Hi Brad,

Yes, very much still here and active."

Kier posted 3 times in 2.5 years, last post was more than a year ago.

I don't call that "active"

 

[Suzanne O]

Kier posted 3 times in 2.5 years, last post was more than a year ago.

I don't call that "active"

He still moderates this forum.

 

[ES Dev Team]

The key point here is with enough consultation with a frontier AI model, large improvement could be made quickly. Failing to start using AI bring xenforo into the modern era I believe is a big problem.

What do you mean by modern? in some areas, modern is a four letter word lately


Xenforo has tons of architectural code. It also has a lot of optimization. Both require very high detail work and holistic thinking to get right. Imagine it as the proverbial foundation of a house.

In these cases, the correct amount of detail for a prompt gets very close to just writing the code itself, and a LLM will usually waste your time instead of reduce it. Maybe you will have a LLM write some parts, at best.

Architecture is more of a design task than a programming task. You need to think about the past, present, and future of all the things that will use it. And Xenforo has a huge plugin ecosystem, so that's particularly hard.

It does seem easy to make features for Xenforo with LLMs. The reason is these features have been made before many time; so many aspects of them are in the training data. Also, the internal architecture of a feature is typically isolated and guardrails are well defined.

Fortune 500 companies do a lot of pioneering work, and also have a ton of architecture. A lot of them have had to rehire a lot of developers they laid off because they found over time that LLM results with architecture are quite bad. Nowadays they report AI doing 10-30% of the work.

 

[zeeb0t]

What do you mean by modern? in some areas, modern is a four letter word lately


Xenforo has tons of architectural code. It also has a lot of optimization. Both require very high detail work and holistic thinking to get right. Imagine it as the proverbial foundation of a house.

In these cases, the correct amount of detail for a prompt gets very close to just writing the code itself, and a LLM will usually waste your time instead of reduce it. Maybe you will have a LLM write some parts, at best.

Architecture is more of a design task than a programming task. You need to think about the past, present, and future of all the things that will use it. And Xenforo has a huge plugin ecosystem, so that's particularly hard.

It does seem easy to make features for Xenforo with LLMs. The reason is these features have been made before many time; so many aspects of them are in the training data. Also, the internal architecture of a feature is typically isolated and guardrails are well defined.

Fortune 500 companies do a lot of pioneering work, and also have a ton of architecture. A lot of them have had to rehire a lot of developers they laid off because they found over time that LLM results with architecture are quite bad. Nowadays they report AI doing 10-30% of the work.

I mostly agree with the foundation analogy, but I’d frame the failure mode a bit differently.

I’ve personally found that AI can automate a surprising amount of development work, but only when the input/design is genuinely excellent and the boundaries are tight. The model needs more than “build X.” It needs to understand the expected behaviour, the existing architecture, the extension points, the edge cases, backwards compatibility concerns, and what must not drift.

That’s where strong guard rails matter. A good example is an extensive test suite. Tests effectively become a contract the model has to satisfy, and they help stop it from introducing subtle behaviour changes while “fixing” or “improving” something.

So when AI-generated code is crap, I usually find it’s because one or more of those things are missing: poor specification, weak architectural direction, no meaningful tests, or not enough human review from someone who actually understands the system.

In other words, AI can be extremely useful, but it doesn’t remove the need for architecture. If anything, it makes the quality of the architecture and the precision of the requirements even more important.

On the hiring/firing/rehiring point, my own experience is somewhat similar, but with an important distinction: the people who remain valuable are usually the very senior or especially capable engineers who can direct the work properly.

Not just “use AI,” but break problems down, define the architecture, write the right constraints, review the output, catch subtle regressions, and integrate it safely into a real system.

So it’s less “AI replaces developers” and more “AI heavily amplifies developers who already know what good looks like.”

 

[zappaDPJ]

The key point here is with enough consultation with a frontier AI model, large improvement could be made quickly. Failing to start using AI bring xenforo into the modern era I believe is a big problem.

I think there's a strong argument to utilise AI in such a way that it provides superior functionality e.g. better search to expose existing content relevant to a specific member.

I also think there's an even stronger argument not to use AI to further develop the underlying code. There are too many potential dangers in doing that which in my view are only going to get worse as AI engines become more powerful.

 

[Ivancas]

I don't call that "active"

He was referring to be active at his Youtube channel:

Kier Darby

Synthesizer tutorials and breakdowns, specialising in the Yamaha MONTAGE family of synths, including MONTAGE M and MODX M. I’m the lead developer for XenForo.

www.youtube.com

Plus he is active on his community and will be pretty busy on the following weekends:

Apologies for loss of service

Sorry to everyone who tried to connect in the last 24 hours and was unable to do so - I moved the domain registration to a more reliable provider but some of the DNS records didn’t migrate properly. Hopefully all sorted now. Look out for some exciting new additions to KDx in the coming weeks -...

kdx.kierdarby.com

 

[ES Dev Team]

I’ve personally found that AI can automate a surprising amount of development work, but only when the input/design is genuinely excellent and the boundaries are tight. The model needs more than “build X.” It needs to understand the expected behaviour, the existing architecture, the extension points, the edge cases, backwards compatibility concerns, and what must not drift.

Yep

That’s where strong guard rails matter. A good example is an extensive test suite. Tests effectively become a contract the model has to satisfy, and they help stop it from introducing subtle behaviour changes while “fixing” or “improving” something.

A big problem with doing something substantially new, complex, and with some other hard requirements on top, is that you find design problems while you are in the process, and adjust along the way.

The problem is the limit of the human brain to visualize and calculate/track sequences of events. And few programmers have an Einstein brain ( unusually high capacity for visualization )

Without being in the process, information on what the actual design should be is missed, and the LLM will generate a nice time-waster for you.
..and the tests you wrote? those turn out to be wrong.

I don't have an einstein brain, so i design the hard parts, then iterate until it's right.
In that case, LLM as a prototyping tool is useful since you can quickly eliminate a few hypothesis and get closer to an ideal one.

The LLM may still write parts of my final product, which are often refactored for shortness and efficiency.
That's because if the thing is super important, the code should be written in a way that ensures future maintainability: simplify it down to the minimum, comment the business purposes of things, name variables extremely clearly, and design flexibility for future additions into it the best you can.

For the work i do, i don't find LLMs super useful. Not because we have the wrong harness, process, or context feeding, but because the processes that really speed you up, do so by front-loading, back-loading, or otherwise removing large portions of the feedback loop that would naturally occur as you built something by hand.

So when i'm using LLMs it's more of a case where we're a tag team, and one of us may be doing more work than the other, but the labor distribution is closer to 50/50

So it’s less “AI replaces developers” and more “AI heavily amplifies developers who already know what good looks like.”

Maybe we could use an asterisk.
For some kinds of code or design tasks, AI might just lightly amplify a good developer's capability. or not at all.

I keep running into these areas and am frustrated often.

To roll it back a bit, i think Xenforo might be one of those projects where in lots of areas, you need high attention to detail, and you might settle on 'human in the loop' processes, IE using a tool like CLine in one's IDE.

I'm sure that if you're building features for Xenforo that don't have much external dependencies, and especially if you are using well designed 'skills' is where your win/loss ratio looks pretty good, no?

 

[zeeb0t]

Yep



A big problem with doing something substantially new, complex, and with some other hard requirements on top, is that you find design problems while you are in the process, and adjust along the way.

The problem is the limit of the human brain to visualize and calculate/track sequences of events. And few programmers have an Einstein brain ( unusually high capacity for visualization )

Without being in the process, information on what the actual design should be is missed, and the LLM will generate a nice time-waster for you.
..and the tests you wrote? those turn out to be wrong.

I don't have an einstein brain, so i design the hard parts, then iterate until it's right.
In that case, LLM as a prototyping tool is useful since you can quickly eliminate a few hypothesis and get closer to an ideal one.

The LLM may still write parts of my final product, which are often refactored for shortness and efficiency.
That's because if the thing is super important, the code should be written in a way that ensures future maintainability: simplify it down to the minimum, comment the business purposes of things, name variables extremely clearly, and design flexibility for future additions into it the best you can.

For the work i do, i don't find LLMs super useful. Not because we have the wrong harness, process, or context feeding, but because the processes that really speed you up, do so by front-loading, back-loading, or otherwise removing large portions of the feedback loop that would naturally occur as you built something by hand.

So when i'm using LLMs it's more of a case where we're a tag team, and one of us may be doing more work than the other, but the labor distribution is closer to 50/50



Maybe we could use an asterisk.
For some kinds of code or design tasks, AI might just lightly amplify a good developer's capability. or not at all.

I keep running into these areas and am frustrated often.

View attachment 341079

To roll it back a bit, i think Xenforo might be one of those projects where in lots of areas, you need high attention to detail, and you might settle on 'human in the loop' processes, IE using a tool like CLine in one's IDE.

I'm sure that if you're building features for Xenforo that don't have much external dependencies, and especially if you are using well designed 'skills' is where your win/loss ratio looks pretty good, no?

Yeah, I think that’s fair.

By “modern” I don’t mean chasing trends or rewriting things for the sake of it. I mean better tooling, clearer contracts, tests where they actually make sense, and safer ways to make changes.

I also agree that hard design work is often discovered during the process. You find bad assumptions, edge cases, and awkward interactions as you build. In that situation, AI can absolutely generate something that looks convincing but ends up being a time-waster.

So maybe the asterisk is:

AI heavily amplifies good developers when the problem is bounded, the constraints are clear, and the feedback loop is tight.

For isolated XenForo features or well-defined extension points, I think the win/loss ratio can be very good. For deep architecture, compatibility-sensitive work, or anything touching the wider plugin ecosystem, I agree the value drops sharply.

 

[Atakan KOC]

I’ve read what was written, and in my view, there’s a slowdown across almost all forum sites. The main reason seems to be that the younger generation isn’t really used to forum systems. But then again, Reddit is technically a forum too, and everyone loves it. So clearly, we went wrong somewhere — the question is, where?

They mentioned programming with AI. Fine, let them do it, but the results speak for themselves: structures full of repetitive, unnecessary lines, hard to debug, and overall messy.

As for XenForo, there’s nothing negative to say. It delivers exactly what you’d expect from a forum platform. Of course, it needs updates and renewal, but that’s natural.

Now, about developers: I’ve been around forums for 20 years, and honestly, I’ve never once spoken directly with vBulletin or XenForo developers. Nor did they ever reach out. Maybe they don’t like talking, or maybe they just didn’t see the need. Either way, I personally love communication, and I wasn’t upset that they didn’t contact me. Since I’ve spoken with developers from other sites, I’d say XenForo’s team really needs to improve their public relations. I’m not saying this as criticism, but sincerely, from experience.

 

[Suzanne O]

I've been known to experiment on all those free forum software and come back to xenforo with different ideas.
It's all ok to do this because then you can take ideas from those forums and use them on xenforo.

Also i'm about to get more and more involved with my own forum.
I have ideas with that.
Sure my forum has a small membership base because it's clear that from joining other people's forums i've had nothing but trolls follow me around everywhere.
This is where i feel sorry for @BIG LLC and his mate @Jim Boy who have had to deal with some nasty nasty stuff on their site that gets a bit personal.

Even on here i've had a few trolls show up.
It doesn't stop me from helping others out on here.

As for the lack of communication i think the xf developers do a fine job both with communicating on here when someone's done something wrong and have asked for help on the forum instead of getting a ticket submitted that would be fixed up later that day by a member of Xenforo's staff.

 

[BrettC]

I’ve read what was written, and in my view, there’s a slowdown across almost all forum sites. The main reason seems to be that the younger generation isn’t really used to forum systems. But then again, Reddit is technically a forum too, and everyone loves it. So clearly, we went wrong somewhere — the question is, where?

They mentioned programming with AI. Fine, let them do it, but the results speak for themselves: structures full of repetitive, unnecessary lines, hard to debug, and overall messy.

As for XenForo, there’s nothing negative to say. It delivers exactly what you’d expect from a forum platform. Of course, it needs updates and renewal, but that’s natural.

Now, about developers: I’ve been around forums for 20 years, and honestly, I’ve never once spoken directly with vBulletin or XenForo developers. Nor did they ever reach out. Maybe they don’t like talking, or maybe they just didn’t see the need. Either way, I personally love communication, and I wasn’t upset that they didn’t contact me. Since I’ve spoken with developers from other sites, I’d say XenForo’s team really needs to improve their public relations. I’m not saying this as criticism, but sincerely, from experience.

I may be able to answer this to an extent...

At the core, where XenForo (and other forum vendors) "went wrong" was user controllability and user flexibility. Forums/discussion boards are traditionally 'locked in' to a single topic type and users do not have a say in what else can be created on it. For example, a Ford Explorer forum is literally limited down to Ford Explorer discussions and some secondary non-explorer discussions.

The second issue in addition to user controllability is then extended to the multi-community front. Reddit fits that bill, Facebook even fits that bill. Why would a user go to different websites to try and find what they want when they can just search a term on a social media platform to find a community sitting there and readily available? The traditional discussion board does not have that 'social aspect' that other mediums have.

Then a third issue has begun to task grasp over the past decade in particular: highly curated 'feeds' to the user (aka: the dopamine addiction hits). TikTok (a rich media platform) is exceptionally good at figuring out what to target users with so that they keep coming back for more and more. YouTube, once their algos get an idea of what you like to watch, will also begin to slam you with a ton of curated content to your tastes. Discussion boards simply do not do this. You have an overall discussion board, categories to that board, forums within said categories, and the threads contained there within. It's all text, there's nothing dynamic and rich about it. There's nothing to curate to a user to keep them coming back. Gamification of a forum does not equate to curated content/feeds.

Again, this issue/problem is not XenForo related, but a 'sign of the times'. Pardon the soap boxing for the next bit. From what I have observed for the past decade, the younger generations (Gen Z and Gen A) do not want to take the time and search, let alone browse a forum for 30+ minutes to look for whatever answer, or to banter with friends. They want the information now, or whatever they can get their hands on via social media platforms. I've seen more and more young people search on TikTok for an answer to their question, and in some cases getting borderline 'brainrot videos' that might have an answer to their question, but is masqueraded as some ad/sponsor or blatantly wrong information. They're not even using Google, or any search engine or website, but a mobile app to get their answers. This is so far detached from a discussion board that it's quite honestly aggravating.

Then there's the fourth issue, and this is something that is of great concern to me: Live/real-time rich media platforms, namely, Discord. The platform is used by all ages, and an increasing amount used by very young teens (truth be told, a scary amount). Most of these younger teens don't even know what an actual discussion board is outside of the Discord built-in forum feature. Why even consider using a discussion board when you can just search on the Discord platform for a server that suites your needs? Didn't find it on discord? Load up your social app of choice on your phone/tablet!

 

[smallwheels]

Brilliant analysis that I fully agree with. Two things to add from my perspective:

the younger generations (Gen Z and Gen A) do not want to take the time and search, let alone browse a forum for 30+ minutes to look for whatever answer, or to banter with friends.

This is not limited to Gen Z and Gen A, despite being more common there. Even many older ones like Gen X or older today rather watch a lengthy 20min YouTube video of dubious quality by an influencer person then do a forum search or simply read 2 minutes in an existing pdf or a manual they have lying on their desk. And often blindly believe what is said in these videos rather than what is said in the official manufacturer documentation.
On top of that I start to see the effects of AI answers that are now automatically blended within Google search results with the same pattern: People just read those automatically generated AI answers (that are often wrong) and blindly believe what's written there. This - until now - seems rather to affect the older generation.

One aspect you missed is the use of mobile devices: In general, forums heavily depend from using the interface and writing somewhat excessively. The "old" internet culture of people actively communicating and creating. Today, we have more of a consuming habit, creating has gone down massively. For one this is a cultural thing and it is fostered by platforms like TikTok, Instagram and Youtube where most people only consume and maybe leave a like or a short comment. Secondly it is also fostered by the massive rise in using mobile devices like phones that are not best suited for writing longer texts or doing complex searches. Plus forum software (including XF) is not very well suited for use on mobile phones, no matter that there is this cheapish PWA available.

 

[smallwheels]

But then again, Reddit is technically a forum too, and everyone loves it. So clearly, we went wrong somewhere — the question is, where?

I wouldn't say we went wrong. Needs and - even more important - habits have changed dramatically, as @BrettC outlined. This plus a dedicated app for Reddit that opens a universe of "forums" for any given topic with a single account and in a single place and is suited for use on mobile phones and easy to consume does the trick. The options on reddit are rather limited (i.e. impossible to post pictures apart from the start posting of a thread) and the overall quality level is rather dubious in many subreddits - but it is fast and easy to consume and today this seems to count more for many than quality of information or real communities (instead of superficial ones).

Classical forums can't solve this "one access to many topics" thing, so I don't think we (as in forums) did a wrong turn - it is simply a different thing that forums cannot offer.

 

[AjayJunkies]

the younger generations (Gen Z and Gen A) do not want to take the time and search, let alone browse a forum for 30+ minutes to look for whatever answer,

Nobody* wants to waste their time and search

Fixed that for us

This is not directed at anyone, and I see this sentiment over and over again among forum owners, but it's the height of hubris among forum owners to think that it's an admirable quality for users to have to trawl through extraneous content, as if somehow that's an intentional feature and not a legacy design.

The rest of the modern web is getting faster, smarter, and more personalized. The question to ask: are forums also getting faster, smarter, and more personalized?

 

[smallwheels]

The rest of the modern web is getting faster, smarter, and more personalized. The question to ask: are forums also getting faster, smarter, and more personalized?

Good question and it would probably be good, if forums (especially XF ) would offer a step into that direction as well. On the other hand the blank copying of how modern social media works could have the opposite effect than many vocal people on here assume: Social media is shrinking and the reasons for it lay deep in how it works. This trend is a chance for forums to take up the momentum.

Just a couple of days ago incognito published the results of a study about social media usage and they are very interesting. They asked 1000 Americans, born between 1946 and 2012, if and how their social media habits have changed over the last five years and why. These are the results (which the posted using the headline The great digital fatigue: How digital burnout is changing social media use)

People are quietly withdrawing from active participation in social media. While they may not be deleting accounts en masse, many are:

• posting less;
• limiting who sees their content;
• deleting stress-inducing apps;
• feeling that maintaining an online presence is work;
• considering leaving because of mental health, polarization, or privacy.

The internet and the connectivity it facilitates have shifted from being focused on interpersonal communication to being sources of news, spaces for political discussions, and sources of algorithmically guided entertainment.

Key insights:​

• People are becoming quieter and more guarded online: 55% report posting less now than they did five years ago.
• 53% have become stricter about who can see their posts.
• Nearly half (47%) of respondents have deleted a social media or messaging app because of stress or anxiety. Rates are higher among younger respondents (61% of Millennials and 56% of Gen Z respondents).
• 44% agree political content and polarization make them want to withdraw from social media.
• Gen Z and Millennials are more likely to cite mental health as a reason for deleting their accounts, in contrast with 25% of Gen X respondents and 12% of Baby Boomers.
• More than half of the respondents state that maintaining an online presence feels like work, rising to 60% among Gen Z.
• Disconnecting creates conflicting emotions: peacefulness is the most common response (27%), but more than one in five also experience anxiety. Relaxation comes third at 21%.

The clearest sign of digital withdrawal is actually not mass account deletion, but a decline in active sharing. The survey was used to investigate how people are reacting to the realities of a dynamically evolving digital landscape. Starting with their posting and content-sharing habits.

Peacefulness was the most commonly reported feeling among respondents who had gone an extended period without checking their messages, cited by 27% of survey participants. Anxiety followed at 22%, while 21% reported feeling relaxed.
he major trend identified by Incogni’s researchers is a withdrawal from outgoing communication, with 55% of respondents posting less now than they did 5 years ago, while 53% have become more selective about who can see their posts.

Gen Z was the only generation to go against this trend. 16% indicated that they now post more frequently than they did 5 years ago, while 13% reported that they have become less selective about their audience.

Overall, however, the dominant direction is toward quieter, more controlled participation: fewer posts, smaller audiences, and greater caution over what is shared.

Their final conclusion:

online engagement apparently used to feel like connection to users and was presumably fun for many people.

Now, the majority of users indicate that maintaining an online presence feels like work, nearly half report having deleted social media apps because of the stress and anxiety they caused, and the most common emotional response to avoiding social media and similar apps is a sense of peace.

From algorithms that demand both attention (from users) and subservience (from creators) to a rising tide of AI slop that’s already displaced all too much legitimate content, it’s little wonder that people are limiting what they share while generally withdrawing from online life. The more pertinent question becomes what can be done to reverse course.

So we should probably be careful what we wish for the direction of forum software...

 

[Wing]

God question and it would probably be good, if forums (especially XF ) would offer a step into that direction as well. On the other hand the blank copying of how modern social media works could have the opposite effect as many vocal people on here assume: Social media is shrinking and the reasons for it lay deep in how it works. This trend is a chance for forums to take up the momentum.

Just a couple of days ago incognito published the results of a study about social media usage and they are very interesting. They asked 1000 Americans, born between 1946 and 2012, if and how their social media habits have changed over the last five years and why. These are the results (which the posted using the headline The great digital fatigue: How digital burnout is changing social media use)


View attachment 341428




View attachment 341431

View attachment 341430View attachment 341429
View attachment 341432

Their final conclusion:


So we should probably be careful what we wish for the direction of forum software...

The future of AI bot written responses and articles everywhere you look - is not going to help at all in this regard either…

 

[zeeb0t]

Life was a lot easier pre-social media. Social media, IMO, enables a sickness present in most of our human minds. Personalised echo chambers. We can't help but be attracted to it. AI isn't much better - often affirming our beliefs, however twisted. It takes skill to prompt it in such ways that your intention is neutral, thus retrieving a... hopefully... more balanced / nuanced response.

For all its "failures" - forums, the old internet, hell - the local town square - a hazard of information, craziness, etc. at least helps keep people centred, because it forces the rest of the world in.

Perhaps the real loss isn’t simply “pre-social media life.” It’s the loss of common informational spaces - places where everyone enters through roughly the same door and has to deal with the same unruly crowd.

 

[Wing]

Life was a lot easier pre-social media. Social media, IMO, enables a sickness present in most of our human minds. Personalised echo chambers. We can't help but be attracted to it. AI isn't much better - often affirming our beliefs, however twisted. It takes skill to prompt it in such ways that your intention is neutral, thus retrieving a... hopefully... more balanced / nuanced response.

For all its "failures" - forums, the old internet, hell - the local town square - a hazard of information, craziness, etc. at least helps keep people centred, because it forces the rest of the world in.

Perhaps the real loss isn’t simply “pre-social media life.” It’s the loss of common informational spaces - places where everyone enters through roughly the same door and has to deal with the same unruly crowd.

Ask yourself if reading 90% AI bot written content and messages sounds appealing to you? I know to me, that is flat out depressing and "dystopian nightmare" feeling. Most of us enjoy still interacting... or atleast believing we are still interacting with actual human beings. I would imagine the future generations will have the same feelings, yet AI is going to take over in huge waves by 2050.

Not liking the outlook for the internet one bit when it comes to genuine human interactions. I predict things may become more "silo'd" again where real humans assemble in small places to retain some of that genuine interaction, but the mainstream internet will be largely AI Bot generated content.

 

[ES Dev Team]

I'm keeping my forum going as long as possible because it's one of the few AI free and social media crap free zones left on the internet.

Meta is getting pummelled with lawsuits and regulations so it's not a far fetched idea that forums may get an increase because they get forced to cut some of the addictive factors.

My current mood re: social media companies