Mindark’s goal of becoming first game studio almost entirely operated by AI

[T16-5 MADMAwith nastya]

Biased opinion; I am a software engineer working in applied generative AI, and have a good perspective of what the current state of the art is capable of. I'll be the outlier to say I'm cautiously optimistic that leveraging AI is the right move for MA.

Balancing: The game is far too large, there are a vast number of items, and the player dynamics are far too complex for a human to properly address balancing in a sustainable and unbiased way. Cutting-edge foundational reasoning models have become very good at problems like this, and as long as MA has properly tuned them and there is sufficient human oversight, I think this will be a good path forward.

Content/coding: I don't know any engineer that isn't leveraging agentic coding now. Latest models are highly competent, more competent than most junior devs, and it's a massive increase to productivity when used correctly. The key, of course, is again human oversight. Our agents aren't just building and deploying whatever they want, they create PRs for human review. A lot of players have complained about the slow pace of development, this is a very effective way of increasing the cadence.

I'm optimistic because I know the technology itself is more than capable. The cautious bit is how well MA is able to apply it.

As for the article, I think it's marketing fluff.

your comment has me revisiting my own thinking when it comes to AI

thank you

 

[endermigne]

Content/coding: I don't know any engineer that isn't leveraging agentic coding now. Latest models are highly competent, more competent than most junior devs, and it's a massive increase to productivity when used correctly. The key, of course, is again human oversight.

Well, from coding I've seen it often go like this (and let me specifically say that I do not refer to MindArk, but coding in general):

1. Developer writes a prompt to specify what the AI is supposed to generate, i.e. what problem needs to be solved
2. Developer copy&pastes the answer into the code

Fine, so what's missing? It's missing the part where the developer reads and understands the output of the AI. Did the AI understand what I want? Does the result solve the problem(s)? Does it meet the coding guidelines and quality standards? This may lead to modify the prompt to get a better result or manual tinkering. So this should be an iterative thing.

I don't know how it's used for content stuff and game designing, but I guess - just as with developers - some creators/designers are better than others.

 

[Lykke]

Well, from coding I've seen it often go like this (and let me specifically say that I do not refer to MindArk, but coding in general):

1. Developer writes a prompt to specify what the AI is supposed to generate, i.e. what problem needs to be solved
2. Developer copy&pastes the answer into the code

Fine, so what's missing? It's missing the part where the developer reads and understands the output of the AI. Did the AI understand what I want? Does the result solve the problem(s)? Does it meet the coding guidelines and quality standards? This may lead to modify the prompt to get a better result or manual tinkering. So this should be an iterative thing.

I don't know how it's used for content stuff and game designing, but I guess - just as with developers - some creators/designers are better than others.

A good point and how I experienced it as well. From what I understand, AI can indeed generate useful code especially for specific, well-defined tasks. But where it still struggles is in understanding the full context of a larger codebase.

So the AI might code / produce something that works isolated, but when you integrate the code into an existing system, where it needs to connect with other functions, and follow a logic, it can often get tricky. I know from experience, as I've been struggling with getting AI to help me out with som PHP coding - and it was clear, that it could code it isolated but not get it to work with the rest (despite feeding it with the rest of the connected PHP files).

AI doesn’t (yet) fully get the “big picture” the way a human developer does. So you still need someone to interpret, adapt, and validate the output. It’s not plug-and-play but more a collaboration between human and machine, where prompts evolve and so does the solution.

It's a bit like with writing or design where the qualty depends not just on the tool but on the person using it.

 

[bicckuni]

I hope one day all humans get replaced by AI, that's gonna be super duper

 

[jetsina]

...and are you hoping to experience that?
What gets me is the stuff that is blatantly false in terms of relevance or difficulty, e.g. virtual tps. Wow, a goto function!
Also the stupidity of claims. It's like a restaurant boasting they have invented forks and spoons - a ground-breaking innovation that lets humans move food to their mouths...
Then there's the maths of millions of transactions per minute. Even if everything counts, whether shots, crafting or mining claims, as well as actual trades, the average per player is going to be well short of 1 per second, so the average number of players needs to be well above 17k-ish. Not even with bots do we get near that.
Can a human please go over what either an AI or another human has written in that article please?

 

[detritus2.0]

Fine, so what's missing? It's missing the part where the developer reads and understands the output of the AI. Did the AI understand what I want? Does the result solve the problem(s)? Does it meet the coding guidelines and quality standards? This may lead to modify the prompt to get a better result or manual tinkering. So this should be an iterative thing.

Yes, for sure it's an iterative process. There's a lot you can do upfront to get better initial results: ensuring proper context, setting up good rules, custom mcp servers for effective data ingress and tool usage. But even the best models still get it wrong, quite often. That's why code review, feedback, and iteration are critical to task success.

I think AI is often abused, particularly in software development, by people who don't understand that in its current state it can't be trusted to build solutions whole cloth without supervision. So you have people who aren't skilled developers themselves, using it to generate garbage, and not knowing any better. As long as the functional goals are met, the user doesn't know how insecure, inefficient, or unmaintainable the code might be.

The good news is, the models are getting better at an alarming rate. I think we will, maybe sooner rather than later, see them become highly effective even in unskilled hands. But we definitely aren't there today.

 

[Timp44]

Imho, After making many disastrous decisions and changes over the last couple of years, the player base has grown tired of the endless BS from Mindark.
They got us playing and paying for future rewards, and when the time came to deal out the rewards, Mindark realised that giving the rewards would slow down/reduce deposits, and Mindark desperately needs our money each day to keep Entropia running.
Unreal must be a long way off, they hardly talk about it now, they are putting all their chips on Ai to fix things, nobody wants it and nobody believes it.
The problem is, they are gambling with our money!

Mindark, please stop.

Why do you think the sudden release of new starter items? They needed capital investment with stock down 90% and financials in rough place.

 

[T16-5 MADMAwith nastya]

Why do you think the sudden release of new starter items?

To lower the cost of play, and it worked wonders.

 

[J4ck]

But these are lies....

How many active players are there? 5-6k or something I believe I saw a while ago?

If evenly spread out on the hours of the day, assuming four hours per player on average per day that makes it at most 1k players online at any given moment. 1k players are not making 60+ million transactions an hour. In order to even get close to 1M transactions a minute we would need to count loot and have at least 6M players. Unless every shot is a transaction in which case we still need a ton more players than there are.

I don't get it. Obvious lies be obvious. This can easily be verified by browsing the forum for a few minutes. As a new player that would be off putting, as an investor that would be laughable.

Perhaps each shot/drop/click is counted as a transaction...

 

[Evey]

Perhaps each shot/drop/click is counted as a transaction...

Loot instance and it would make sense actually...

 

[Love Banjo]

Just another ai generated media stunt not much to see here

 

[Lykke]

Perhaps each shot/drop/click is counted as a transaction...

That's most likely true. A transaction is an action and counts in all activities a player has: Trade - all items and ped exhanges, all auction activities, all events / actions connected to gaining loot, crafting, skill-ups, all ped-operations like deposits, transfers, every click with a gun, etc.So basicly all in game actions made by a player. Still 60 mio transactions per/hour does seem high for a 1K-1.5K average amount of players / day. However, every little click in a terminal counts too - and some players are crafting enormous amounts of clicks / transactions every day.

 

[J4ck]

That's most likely true. A transaction is an action and counts in all activities a player has: Trade - all items and ped exhanges, all auction activities, all events / actions connected to gaining loot, crafting, skill-ups, all ped-operations like deposits, transfers, every click with a gun, etc.So basicly all in game actions made by a player. Still 60 mio transactions per/hour does seem high for a 1K-1.5K average amount of players / day. However, every little click in a terminal counts too - and some players are crafting enormous amounts of clicks / transactions every day.

Perhaps even multiple transactions per click: ammo, decay (gun with deveral attachments etc). Could add up really fast that way.

 

[Lykke]

Perhaps even multiple transactions per click: ammo, decay (gun with deveral attachments etc). Could add up really fast that way.

yes, that sounds possible indeed

 

[TeesMaarKhan]

I knew something shitty was going on in the background for the last 4 or 5 months because of blatant return shut down, it becomes clear. We have a bot they call AI that's been programmed to shut returns down when this happens when this triggers and so on.

 

[San]

I think what rubs people the wrong way about AI is not that AI is now in the toolbox and gets applied to tasks which were tedious before, even for highly skilled humans. It is terms like "AI-led" which seem to imply that AI is put at the top, not at the bottom where it belongs to put an end to sweatshop-style drudgery. If players get told that their content is no longer made by creative human beings, but by machines, then it is no longer perceived as entertainment, but as something else I'm not sure yet how to name. You just know you're no longer spending money to reward people whose designs you appreciate. From employing AI to letting yourself be led by it...

I don't know if this is actually true at this time, or if it's a vision for the near future, or if it's just total baloney. MA's press releases have been in the latter category too often to take any of them seriously before any changes are seen on the ground. But I'm getting the impression that at the very top, if there is anyone of the game's original creators left, there must be a growing sense of wanting to retire, not only for all the right reasons due to the natural course of life, but for the realisation that the once endearing thing they created has become a monster through the kind of people that its original concept attracted. A classic Faustian conundrum -- "the spirits I summoned, I no longer control." Except we now have AI to keep them spinning in endless loops, while we quietly slip out of the door.

 

[Katla]

EU has slowly but surely turned into an idle game. In that regard it makes sense if MA want to make it the first MMO game in the world solely played by software. They should consider a new name though- Softropia Serververse

 

[TeesMaarKhan]

Edit....just Not worth typing here any more...

 

[jetsina]

Had the same feeling of not worth it for longer-term commentary, and hardly for short-term either. Still, some great commentators haven't given up completely yet, so kinda responding with 2 pecs here and there.

 

[Evey]

The most important thing about tis article to note s that it's not for us.

MA communication is for one of the 3 major directions: playerbase, investors and general public.

For each of the groups, communication for the other groups seems weird at times. Investors won't give a damn about efficiency, dpp, lag, stuff like that, players won't really be very interested in IR releases and general public doesn't understand very well so more flashy and buzzy words are used...

In each group there will also be a wide spectrum of people who understand technology at different levels so they will react differently. So the news released is not for us... but it annoys us, mostly because, to us, it sounds like less human touch is involved and as san said, we get the feeling there's no more creativity involved and I think the game would not have survived this much without passionate teams working on the product.


When it comes to AI we're in a transition point, some people adopted it early, some try to understand it and some can't so they fear it or despise it for what it is but reality is that it's just a tool and it's here, now, present in our lives and it's not going anywhere. Smart businesses already incorporated it, others are in the process of doing so, but in communication there should be a different approach and should keep in mind the people's concerns about it and aim to communicate for the entire spectrum of understanding and type of adopters. Maybe when doing such a release, different announcements should go for all 3 targets at the same time to not cause too much disturbance.

As always, when it comes to communication, MA has much to improve... When MA? WHEN?

 

[GeorgeSkywalker]

too much disturbance.

As always, when it comes to communication, MA has much to improve... When MA? WHEN?

Probably never!

They'll just go into burying head in sand mode or we commonly recognise as silence.

 

[Lykke]

As always, when it comes to communication, MA has much to improve... When MA? WHEN?

Yep. Communication isn't just about the occasional press realse but about consistent and clear messaging across the board to the player base, investors, new and potential players, to partners and the wider gaming industry. Only a few - this being Planet Partners - have been able to make good communication and advertising.

It's not limited to the press either, but should include regular updates on official websites, LinkedIn, company profiles, dev blogs, and other relevant platforms. The broader communication strategy feels a bit lacking - and it's something that definitely could make a huge difference in building trust, excitememt and visibility.

Hopefully this will improve in future, perhaps even with the help of AI No doubt a more transparent and proactive approach could benefit everyone involved.

 

[Timp44]

To lower the cost of play, and it worked wonders.

The only ones it lowered the costs for were the people at the bottom, you bought that hook, link, and sinker. You think they care about you being able to play cheaper? They care about the revenues and are praying that the starter packs will hold higher retention of players.

MA lost a lot of money last quarter. You'll see the influx of new cash on the balance sheets come this next quarters report.

 

[Droops]

So you have people who aren't skilled developers themselves, using it to generate garbage, and not knowing any better. As long as the functional goals are met, the user doesn't know how insecure, inefficient, or unmaintainable the code might be.

The good news is, the models are getting better at an alarming rate. I think we will, maybe sooner rather than later, see them become highly effective even in unskilled hands. But we definitely aren't there today.

 

[Evey]

Do these new AI agents work during summers?

 

[ProActive Mango]

AI is the future. Adapt or Die.

 

[exsolvo]

Well, that actually explains it, if they are using AI to write the code, then it's no wonder that half the game breaks on every release, as that is exactly my experience with every AI-agent i've tried to do coding with.

I.e:

1. Write as detaled a prompt as you can, describing the "issue", give clear context.
2. Watch AI create one or more solutions
3. See which solution seems to be the lesser evil as per what it breaks
4. Prompt over and over to clearly have the AI _not_ break other things that you see
5. Realize after countless hours that if you just did it yourself, you would have been done a long time ago.

No AI on the market today (as far as i've seen, and i've used pretty much every free version there is, and multiple paid ones) has a good grasp on:

1. Staying within spec, they all want to change things that are not relevant to the task at hand
2. Remember what ended up NOT working so as to not suggest the same thing again. (And THIS seems to be what MA may be strugling with, because every fix they do seems to break something else).
4. Feed-back loops, i.e after a certain number of "That doesn't work either, try something else" AI's tend to get stuck in a feed-back loop just regurgitating failed attempts over and over.

So yeah, In todays market, AI is a bit too overhyped imho, it's more a fancy search engine than an ACTUAL code companion or even less full fledged senior developer.

 

[TeesMaarKhan]

Seems ABIT like

MA = is the car owner
AI = parts shop

MA to AI = Hi I need a specific mods for my car to work better.

AI= what car is it?
MA = That I'm not willing to reveal I just need something to make it faster and better than it is.

AI = okay let me see what I can come up with, runs off to the back, comes back with what ever it could come up with.

MA = goes away and attaches all these non compatible parts no ECU coding no nothing, hey they are attached start her up...

BOOooOom

And that's why everything is always fkd.

Rince and repeat.

 

[mihaimnb]

Given the scale of botting, I reckon we're there already, MA just playing catchup

Totally agree here...most players are not even playing just jokes, even some ubers

 

[GeorgeSkywalker]

Totally agree here...most players are not even playing just jokes, even some ubers

Even that they can't get right. I mean everyone knows most people bot, but no one is allowed to admit it though.

If you are going to go down that route better to do it properly. This marketed properly would attract a lot of new players e.g. as aids to grinding.