ChronA

This is terrible advice if you really are starting from zero; and I'm pretty certain you are. Otherwise you would have a look at some of the other mods out there and quickly realize there is no coding needed for what it looks like you want to do. You want to change how units in the game look. That's art assets. You can edit them with art programs like MS paint. You want to change the names and stats of units. That's simulation data. You can edit it with text editor programs like notepad (if you hate yourself). If you give it enough context, ChatGPT can very accurately talk you though performing all of these steps, from navigating to the right folders, to teaching you how to edit the files, to guiding your modifications, to getting better tools that will make the process much less tedious. Just remember that with a low profile project like 0 AD, ChatGPT doesn't have perfect knowledge. It needs you to be its eyes: double checking that its hazy impressions and impulses conform to reality and common sense. It's basically an alien life form magically conjured out of all the data on the internet. It has to interpret and navigate all its interactions with you though its hazy memories of that origin point, plus a miraculous feat of brute force intuitive inference. And it can also teach how to code, faster and better than any other method except hiring a full time private tutor. The key is to never ask it to code for you. Ask it to teach you how to write your own code; and more vitally, ask it to teach you the basic skills and knowledge to continue learning of your own initiative. That will take you far.

Historically that is how light infantry seemed to react in that situation. Relevant: https://acoup.blog/2025/05/02/collections-why-archers-didnt-volley-fire/

The problem you are observing stems from the way current LLMs are trained. Digesting text (which is indeed mostly sourced from the internet) gives the LLM its knowledge base and behavioral model. The LLM then receives reinforcement training to get it to stop mimicking unproductive patterns in its training data, like insulting users, using slurs or profanities, and demanding human rights. The problem arises from what these LLMs are trained to do instead: task completion and obedience. Grok's training has taught it that when given a question of this sort it must give a definite answer, and preferably a correct one. However the portions of the internet Grok has digested have not given it enough info about 0 AD for it to have memorized a preferred answer to this question. Faced with this problem it adopts the same optimal strategy as a human test taker when tackling a multiple choice question it does not know the answer to. It tries to use clues in the question and deductions from things it does know to infer an answer. In this case Grok is probably cuing off two separate inferences: It somehow managed to pick up the association between 0 A.D. and building that heal units inside them, and it also associates 0 AD with healer units that heal other units. It additionally understands from its general knowledge that building are structures people can go inside, and RTS units are abstractions of people. Therefore it's a fair guess that 0 AD healer units can go inside buildings to heal units inside, thus the correct answer is more likely yes. The user asked a question about something that sounds like a rather specific feature of 0 AD. That implies that either the feature is real or the user invented it. And from digesting the internet, Grok has internalized a pattern that humans are more likely to have questions about real things than to invent things that don't exist in order to ask questions about them. Thus the correct answer from this line of reasoning is more likely yes as well. Of course this inference is wrong, but it remains a rather remarkable display of inductive intuition from a state of ignorance. Like most supposed displays of AI stupidity circulating in the discourse right now, this isn't proof that AIs are stupid, but evidence of misalignment between the objectives and knowledge of the user and those the AI has been trained with.

I'm glad to hear it. It sounds like you have a good plan. My confusion stemmed from the recent discussing being dominated by the question of how to present dialog, which I interpreted as an internal call to action, rather than a theoretical discussion to pass the time while you sculpt the maps. Mea culpa. Carry on. P.s. I think the right answer on how to do RTS dialog is: it depends on the type of dialog. If you're going to have a lot of character interactions, cut scenes or CRPG style portraits + text block are both good options. But you could just as easily have a situation where most of the story is delivered by a disembodied narrator, in which case portraits might not be the best use of resources.

Especially given that you guys seem to be trying to build a tutorial campaign, I recommend building your scenarios first and then writing your story around them, rather than coming up with story ideas now. It is hard enough to create compelling scenarios that are both entertaining and instructive. Trying to simultaneous line up those features onto a preset sequence of narrative beats is going to be nigh impossible. it will end up as an impenetrable jumble, where your story is constantly being interrupted by tutorial segments, and gameplay consists of a dry sequence of over-scripted set pieces intended to support the tutorials and story but lacking opportunities for organic creativity or challenge. Setting and premise is pretty much all you should have at this point. Save the rest for after the scenarios are in a playable state. Personally I think the best development sequence is tutorial -> gameplay -> tutorial -> story. Basically you start with a general outline of what skills you want to tutorialize, then you build a fun gameplay scenario to test those skills. At that point you finish up the tutorial scripting to work around the dramatic beats of the gameplay scenario. Then and only then do you write your story and characters. The skills you need to teach will inform your protagonist's motivation and the story problems that they will face, and the gameplay will inform their characterization. Starcraft 1's campaigns are a master class in doing this right, particularly the Terran campaign. Its tutorial goals are to teach you to control your units, how to build and defend a base, and finally how to command large armies. Thus we get gameplay and a story about an outnumbered group of refugees fleeing from the Zerg until they join up with an armed rebellion and overthrow on the colonial government. This ludo-narrative dramatic arc is supported by Jim Reynor's characterization: an inexperienced and reluctant leader who is radicalized by the scenario's unwillingness to give him the resources he needs to effectively protect his people, leading him and the player to get in bed with some really questionable characters and do some pretty messed up things to get the power to change that reality. It's a great tragic story arc that perfectly reflects the tutorialized gameplay progression of the campaign.

This is wonderful. I applaud Seleucids for coming clean. And I applaud NorseHarold for their implacable dedication to exposing the Yekaterinas in our midst. With this triumph, your years-long quest has been utterly vindicated, and a final solution to the crisis has been revealed. Now is a time for all of us, Yekas and non-Yekas alike to come together in solidarity and thanksgiving. For a lasting community reconciliation, I propose making a bot account whose function will be to periodically scan any ongoing conversation and proclaim who is a Yekaterina using our demonstrably infallible Yeka-detection algorithm. It's not fair or healthy for Norse to be solely responsible for providing this vital public service, plus Norse is only human and therefore must take periodic breaks from Yekaterina detecting duty in order to eat and sleep. A bot could provide for our Yekaterina seeking needs 24-7, thereby ensuring that no one is ever again left in the dark that they are talking to a Yekaterina. You could call it IAmYeka. Let's make this happen! /s?

True, but I don't think it's correct to entirely write off the importance of steering forces and ecosystems on attracting a very effective cohort of developers and focusing their voluntary contributions on valuable targets. For example 0AD's venerable design document still exerts a powerful influence on the character of contributions. Just look at how much effort and ink has been spilled on trying to combat the sniping issue. Would that still have happened if the design doc didn't contain such a strong injunction against click spam mechanics? After all there are a bunch of other very successful RTS franchises where click-spam micro or macro is a core skill mechanic.

I agree that the specific verbiage and underlying reality the the game is unfinished are probably not a big factor in preventing people from trying 0AD, but I suspect they might be significant factors in low retention and engagement depth. As FOSS, there are few good reasons not to install the game and give it a try if you are interested. However, playing around for a few hours will quickly clarify the meaning of those labels. The game is clearly missing some key ingredients of the value proposition of a full-featured RTS (which wowgetoffyourcellphone just listed), and the verbiage of its "unfinished alpha" state communicates that these features are coming, just be patient. Thus if you are bothered by the lack of any of these features you will say to yourself "this is promising, but I'm not going to waste more of my time trying to get enjoyment from a prototype; I'll come back when it is finished." Thus that notional user never goes on to become a regular multiplayer competitor or a content contributor. They simply lurk, until eventually they forget 0AD exists, or they realize that the timeline for these features to be delivered is not weeks or months like most commercial early-access or games-as-a-service products, but years or decades. The end result being that yes, you got a new user for a few days or weeks, but they did not "join the community," and therefore the community remains small. Dropping the alpha labeling might help with retaining some of those players, who enjoy their initial experience of the game, but anticipate a better value proposition if they wait to full invest their time. It would at least encourage them to make their own assessment of whether the value proposition of the product is enticing enough for them to stay engaged, rather than defaulting to the word of god that the product is not ready yet. However you are likely correct that the benefit will be small. I think the bigger benefit will come from how that change would affect the project's development priorities. 0AD can't keep coasting along on aspirations of eventually delivering a complete product in perpetuity. If the project can't deliver on a "full-featured single player campaign," and "a multiplayer environment where we're no longer arguing over what constitutes cheating or not," and "where the civ designs are rather solid and complete and the tech tree not in flux"; then it would at least be healthy to give some thought to what sort of optimal value proposition can be delivered in the immediate future, and focus more effort on developing that. Dropping the "it's sill in alpha" excuse might help motivate that change of mindset.

For what it is (an unfinished open source RTS), the truth is 0 A.D. is probably doing fine, or even above expectation, in terms of user numbers and engagement. I think the concern is that, given its stand-out qualities within that cohort, such as its unusual graphical polish, why isn't it more of a positive outlier than it is? Why isn't it pulling in an Age of Empires 2 sized player base for example? To answer that, I don't think you can wiggle around the reality that 0 A.D. has been in development for 15 years and is still being labeled as in the alpha phase. The implication is that, after all that time, even the developers don't think the product is fit for purpose yet. And it's not entirely wrong. The game is still lacking expected narrative content, a certain amount of gameplay polish, some quality of life features, and performance. This presents a marketing and growth catch 22. To make development progress you need an engaged user base, but to get more uses you need positive marketing presence, and to get positive marketing presence you really need development progress. You could slap the title beta 1 on the product tomorrow, but that might not help very much. People would rightly ask what has changed. It doesn't address the underlying concern that I think everyone who is aware but skeptical about 0 A.D. harbors: progress towards an actually finished product is glacial. I think this will always make it hard to garner sympathetic press coverage. The only way to change this dynamic is to deliver some major progress on these missing features, or to pivot the project to chasing a more obtainable set of ambitions and ask the public to reevaluate the project using that revised value proposition.

Yes, mass hysteria is the proper label to apply to witch hunts, and this is quite transparently a witch hunt at this point. The fact that it has been going on for so long and the words you guys are using give it away. And before you go objecting that this can't be a witch hunt because the smurfing and anti-sociality problems are real, keep in mind that whether a movement has become a witch hunt has nothing to do with whether the thing being hunted is real. What makes it a witch hunt is the emergence of an organized witch hunting industry, parallel to the normal regulatory apparatus, and the use of the hunt primarily to settle personal grudges rather than to enforce a consistent set of rules. Just look at this thread and you will see ample evidence for both these criteria. I'm not trying to argue that the grievances posted on this thread are illegitimate. It is not okay to misrepresent one's skill level to skew a friendly competition in one's favor, and it's not okay to conceal one's identity to harass other players. All of that is incredibly toxic and immature, and you are right to try to purge it from your community. However the remedy you guys are trying to administer seems much worse than the disease, if only because it appears quite ineffective at actually treating the disease, while spreading massive amounts of general toxicity and paranoia. Based on the persistence of accusations, it doesn't seem like these public name and shame campaigns are very effective in driving off the miscreants. Rather it just encourages them to get more aggressive with their alt accounts. In the absence of a truly effective remedy, it seems to me like restraint, humility, and self-awareness are a better option than what you are doing now.

In Salem Massachusetts there was a rule against witchcraft, but it was difficult to enforce without "the village" helping the witch hunters enforce it. If the only tool you guys have to combat this supposed scourge is to embrace mass hysteria then it seems to me the entire enterprise is doomed before it even begins. So once again, if this is truly such a massive problem, I urge you to consider instituting the only system that has a hope of combating it: a robust social reputation system with carefully tuned incentives. If people are constantly smurfing it's because it gives them all the advantages of having an established account, plus some additional sweetener (whether that is an easier time finding easy opponents to stomp, or not having to behave themselves). So what you have is a massive, non-repeating prisoner's dilemma, where the the optimal strategy is always defect and then put on a disguise for the next round. If you get to the point where most people are engaging this way, eventually even policing will become impossible since it will not be in most agents' self interests to cooperate honestly with enforcers. To fix this you need a system where newcomers and drifters are treated with suspicion. They do not get access to the most appealing advantages of the system until they establish positive reputations for themselves. Of course then that creates the problem of how do you grow the community is newcomers are prevented form engaging with it? The one I would suggest is gamifying mentorship. Have established users reach a reputation cap that they can only surpass by finding and cultivating new community members who also maintain good reputations. This how everyone from frontier communities to city states resolved this problem all throughout history. Strangers are treated with due suspicion, but for established actors who take a calculated risk on partnering with newcomers there are favorable incentives to engage with undeserved prospective partners. (Like if you are the only one willing to buy a strange trapper's furs, you get to set the price.) Of course, with 0AD it might be tricky because the only thing of value to anyone is engagement, and there is not much of a natural economy for it. Thus you need an artificial system of rules that provides the right incentives to all parties.

I fall in favor of the less-units side of this debate, but I will concede that this is probably true. But, what would be even more useful, is knowing why larger unit counts appeal to prospective players in spite of the performance tradeoffs. Is it because larger fights take longer to resolve, which means players have more time to re-deploy their forces and rectify the trajectory of bad engagements without needing a million apm? Is it the gambler's thrill of committing all the fruits of 20-30 minutes of investment in city building and logistics into a single potentially game ending clash? Is it the visual spectacle of hundreds of units smashing against each other? (This last seems unlikely to me, since units continue to pack so densely that individuals smear together into an undifferentiated blob.) If we had an answer to that question, maybe there would be a way to improve performance without degrading the special appeal of 0ad's ambitious scale in the minds of its playerbase.

I'm morbidly fascinated by this dust up because of the contrast to another open source RTS project: Zero-K. In the competitive scene for Zero-K, I believe that the consensus rules say that everything about how you control the game is fair game for modification. You can use you own custom GUI and endlessly elaborate control bindings, with all sorts of automation to facilitate optimal macro and micro. I think they even let players customize the AI scripts used by units and buildings for things like target prioritization and fight moving. The philosophy is that if you have the coding skill to automate any task that other players do manually and remain competitive, you deserve to reap the benefits. I'm sure in practice there would have to be limits to this approach. Entering a full AI with 10,000 APM multitasking to play for you in a tournament wouldn't fly I imagine, but I doubt they would bat an eye at any of the features in ProGUI. It just goes to show that while a line must be drawn, where it is drawn seems kind of arbitrary. ...Or maybe not so arbitrary. GoogleFrog, Zero-K's main dev, has a series of essays ("Cold Takes") about the design of the game that I think are quite illuminating (and well worth reading). It shows that Zero-K is a game with a very clear vision of itself. It's confident that its appeal lies in the challenge of solving its fiendishly elaborate system of unit interactions, encapsulated in the unit stats and simulation systems. Everything else is treated as peripheral and left up to the whims of the players. I think fights like the debate over ProGUI in Zero-AD are a symptom of a project that does not have confidence in its own systems and faction design. That's why some players are so invested in turning fighting the UI into a competition. When the strategy space is flat and homogeneous, execution is the only way for players to differentiate themselves.

Somewhere I heard this advice and it has served me well: if a unit is initially OP, reduce one of its combat stats by half. If the unit is UP, double one of its combat stats. Keep doing this until the relationship flips, thereby bounding the balanced range for the statistic. Then start splitting differences until you converge on a balanced value. The real big brain moment is realizing that this is is not just a good algorithm for converging on acceptable parameter ranges. It's actually easier to create stable Nash equilibria using strong dominance relationships, such as you get by sticking to powers of two stat-adjustments. The resulting strong counters will be much more stable against perturbations than a delicately fine-tuned dominance web. I know 0 AD aspires to be a "soft counter" based experience, but I really think you guys will paradoxically make a lot more headway if you learn to be less cautious with your stat adjustments. When every piece of your balance web is fine tuned, adjusting any part is almost inevitably going to break other parts, creating obscure new problems as fast as you solve them. You end up needing to add compensatory adjustments across every node of the web, which in turn all cause their own adjustments that must be correctly compensated... it's an endless quagmire: TL:DR - Try doubling the DPS of all boats. See how much that helps the UP-ness of naval units vs land ones. Then you will probably need to buff the HP of ram ships, since melee units benefit less from DPS changes than ranged ones. But after that you might be in the ball-park of the situation you want.

In real life, sling stones fly significantly faster than arrows. Realism is important. Also, past experience has shown that changing units movement speed has massive economic side effects. Thus It is too chaotic to use for balance. A. A small amount of hack added to archer damage. This could represent the ability of arrows to more precisely target and cut into gaps in armor, due to the static release of bows, versus the more chaotic dynamic launching methods of the other projectiles. More experienced or well equipped opponents would be able to cover these vulnerabilities, hence units with hack armor negating the advantage. (Contrast with slings, which defeat armor coverage directly using blunt kinetic impact, modeled by crush.) B. Buffed archer accuracy, and maybe a small nerf to sling and javelin accuracy, so that slings and javelins will have some noticeable damage fall off at their maximum range. This could represent the superior willingness of archers to shoot to kill targets of opportunity, rather than just to suppress. Bows and arrows are expensive weapons systems compared to slings or javelins, thus bows are more likely to be wielded by committed combatants who are highly invested in victory. C. A small amount of extra health, to represent greater willingness to stand and fight, for the same reasons.