In board games, my campaign of Tidal Blades 2 is in full swing, and we finished scenario 11 out of 18 yesterday. We usually get 2 scenarios played per session, and currently play about twice per month, so we will probably finish this campaign around the end of the year. My other regular board game campaign is currently on hold, as the couple we are playing with just had their first child.

One highlight of my board game life last month was the visit to the Spiel 2025 in Essen, the world's largest tabletop games fair. I was rather pleased with my logistics organization for that, including parking near the fair a day early, as you can't get parking there during the event. But while the organization went well, the event itself left me unsatisfied. It was extremely crowded, and so, in spite of me being there all day, I had very little opportunity to get some game explained to me. That resulted in me buying games based more on a quick look or on what was considered hot by BoardGameGeek. And then the first hot game I actually played after the fair, Covenant, turned out to be not that great, at least not for me. As I can't optimize my fair visit any further, I currently think that I won't be going next year.

When I wasn't away for the fair, my regular board game nights continued. From the games I brought, we played Dune: Imperium and Steam Power. From other people's games I played 7 Wonders Dice and Finspan; while it was interesting to see the evolution of the "X-span" series, I didn't like Finspan very much, and find that the series is moving into the wrong direction, with decreasing player interaction. I might not have been the only one thinking that way, as we finished Finspan in an hour and a half, and would have had time to play another round, but nobody even proposed that.

In video games, I cancelled my PC Game Pass subscription last month. Not really because of the price increase, but also because I won't be needing it much when I play EU5. Before that subscription ran out, I played both Endless Legends 2 and Metaphor: ReFantazio. Both of which I rated "I'm happy I didn't pay for this". Endless Legends 2 looked a mile wide, and ended up being an inch deep. In Metaphor: ReFantazio, I was missing player agency, with both the endless cutscenes and dialogue and the time limit getting in the way of me playing the game the way I wanted. So, yeah, this is exactly what I had the Game Pass subscription for, and I'll probably return at some point, but I'll play EU5 and other Steam games first.

Master of Command covers The Seven Years' War, which was one of the first more global wars, and resulted in the rise of Prussia to a great power. In the game, you can choose between 5 nations, each of which have 3 campaigns (you need to finish the first one to unlock them all), and each campaign has 3 acts. In each act you are moving a single army around a procedurally created map, and interact with map points like settlements or forts. You have only a limited time until you need to face the main enemy on that map, so you need to gain strength by fighting, looting, and various map events. Every battle involves a Total War like real-time fight, which you mostly win by breaking enemy morale. You have usually a lot of infantry, some cavalry, and some artillery. Some fighting is done by lines of infantry firing muskets at each other, but there is also melee combat.

In each act you both grow your current army in quantity, but also in quality. Battles give experience, which allows upgrading of units. You can gain items as loot in battles, from map events, or buy them in cities, and those items give bonuses to your regiments. The scale remains reasonable, with a brigade having up to 4 regiments, and you typically growing from one to three brigades in act I. Learning how to play is easy enough, learning how to play well isn't, and you might actually want to start this one on easy (Lieutenant), or with at least the "Early Campaign Season" bonus modifier active.

Different nations focus on different things in their armies, quality or quantity, infantry or artillery, and so there is some replayability. But this game is also fine to just play one or two campaigns and be done with it, if you don't want to go deeper.

In many of the video games I play, especially in roleplaying games, my options at any given moment are limited, and restricted to the situation I am in. I might be in a dialogue and have dialogue options, I might be in combat and have different attack options, or I might be on an overland map and have different options where to go. But when I am in a dialogue, I can't fire a magic missile at the person I am talking to (unless that is a foreseen dialogue option), and when I am in combat, I can't talk to my enemies. Europa Universalis is different to that. All the buttons of the game are available all the time. The game doesn't tell you that you are now in a dialogue and should pursue diplomatic relations with your neighbors, and while some diplomatic options might be impossible during war, the diplomatic menu never goes away completely.

And unlike most other games, choosing one action doesn't necessarily prevent you from doing something else. In most games, if you choose for example to cast a spell, that is your action for that turn in a turn-based game, or for so many seconds in a real-time game. Even in a real-time game with pause you might at best be able to give a series of commands that will be executed sequentially. In Europa Universalis and similar grand strategy games, you can pause and then give dozens of commands in different parts of your menu.

All the buttons being available from the first moment of the game and simultaneously creates a bit of an accessibility problem. Paradox published a video today called Here's Everything You Need To Know Before You Unpause Europa Universalis V, because there is a whole game you can play to go through all the menus and understand your chosen country before unpausing and letting anything happen for the first time. Different countries have very different starting situations regarding their economy, their population, their estates (power groups), their military, and their diplomatic relations. I will certainly take a good amount of time on my first game on Tuesday to just go through everything and see where I stand, and whether there are actions I need to take early. In EU4 I once played a game as the Teutonic Order, and there is a specific disaster looming over you, which will very much mess up your game unless you know about it and address the problem from the start. For a new player, having all the options and not knowing what is important to do now, and what can wait, can be really overwhelming.

But of course there is also the option of just letting the game run and see what happens, treating it more as a simulation than a game. In the past weeks, EU5 has garnered some criticism based on somebody showing a time-lapsed video of the world without the player doing anything. Some people felt that in that situation, the AI wasn't aggressive enough, and countries didn't expand enough. Especially some nations, like the Ottoman Empire, appeared un-historically weak. For that you need to know that EU4 starts in 1444, at which point the Ottoman Empire is already large, while in EU5, starting in 1337, the Ottomans still have a rather small country. While historically the Ottoman Empire was a huge threat to Christian Europe, sieging Vienna twice in 1529 and 1683, and EU4 reflects that, in EU5 it isn't a given that the Ottomans grow into a large empire at all. In fact, if the player chooses to play Byzantium, they will almost certainly stop the Ottoman Empire from happening.

There are several interesting design questions involved here: One is how much you tune the AI towards aggressive expansion and huge empires. Historically, global empires are rare, world conquest (which is something some EU4 players like to do) never happened, and regional empires had a tendency to break apart over the centuries. The way EU5 AI is programmed now seems to be in line with both history, and the player expectations of how the world would look like after the first 107 years, from the start date of EU5 to the start date of EU4. If in a typical game of EU5 in 1444 for example the Holy Roman Empire would be largely united, or consist only of a handful of countries, that would be both unhistorical and somewhat weird as well.

Which gets us to the second big design question, which is in how far designers just let the simulation evolve countries, or in how far they add scripted events. EU5 starts less than two months before the historic start of the Hundred Years' War and nine years before the Black Death. These are events that will most certainly be scripted to happen in EU5, as will the Protestant Reformation in the 16th century. Other big historical events, like the discovery of the New World, has been left to happen "naturally" in EU4 and EU5, controlled only by naval technology. Which means that I have seen EU4 games in which AI Denmark colonized Canada and North America, rather than the English or the French.

The big problem with scripted events is that players quickly reverse engineer those scripts, and exploit them when playing in a region where they happen. Read or watch any guide on how to play France in EU4, and it will tell you how to deal with the Surrender of Maine or the Burgundian Succession, both of which are important scripted events. I don't even know whether these events will be in EU5, because with over a century of an earlier start, how likely is it that England will hold Maine in 1444, or that the Burgundian ruler dies in 1477 without a male heir? Which gets us back to the rise of certain empires in history, like the Ottomans, or Prussia, which historically involved some element of chance, and thus in a pure simulation have a good probability of not happening at all. In EU4 there was a game mechanic called "lucky nations" (which you could turn off), in which certain nations got bonuses that made it more likely for them to become important and play their historical role, but the devs decided to not do that in EU5.

Fortunately for me, none of this is a problem. I am okay with EU5 as leaning a bit more towards a simulation, and being a bit less game-like. I never liked world conquest gameplay in EU4 anyway, and am okay if it is strongly hindered by problems of large empire stability, as that is historically realistic to me. I'm also okay if certain things in my game don't happen as they did historically, as long as they seem somewhat historically realistic or possible. And I realize that the number of historical events and scripts added to EU5 is very much a function of time, as they take time to program, and much will be added via DLC. That doesn't stop me from playing the game on release, with a bit less of all that.

While it took me months to write my thesis, and years to prepare, modern generative AI can produce a very similar looking book with an afternoon's amount of work of somebody prompting it. It would be equally incomprehensible to most people, but look nicer, and appear to be much better written. Only a detailed examination by a specialist would reveal that the document holds no scientific value at all. By definition, something written by generative AI is either copied from somewhere, or hallucinated. It is work slop, like Deloitte's AI-written report for the Australian government. Looks good, until you verify the data and the citations, at which point the hallucinations become obvious.

Most of my life's work resulted in written documents: My thesis, several scientific publications, countless technical reports, and 32 patents. Arguably I would have loved an AI to help me write the fluffier parts, but the value of those documents is all in my data and my work they report. You can get generative AI to write you a patent for a perpetuum mobile, but that doesn't mean that you'll have a valid invention on your hands. Scientific documents, legal documents, engineering documents, everything where the value lies in facts being reported correctly, can be faked by generative AI. But unless you only let the generative AI write the prose and let the expert write the data and conclusion from those data, the AI document is worthless.

I am pretty sure that all my patents have been verified and checked by an expert, because as an inventor you get the comments from the patent office examiner back. Scientific publications are peer reviewed, and a committee verifies any thesis before awarding a degree. With my technical reports I wrote for internal use at the company I worked for, I am not so sure. Most people probably only read the cover page, and detailed scrutiny by another expert is unlikely. If Deloitte dares to deliver an AI-written report to a government that paid them, I can easily imagine somebody in a company writing his reports with AI without making sure that all the facts are correct. Work slop looks nice, it just doesn't hold any actual value. Just like AI generated images are getting harder and harder to spot directly, AI generated written documents aren't immediately obvious. If somebody got through college with AI-generated essays, why wouldn't he deliver AI generated work slop in a company, when his supervisor asks him to quickly produce some report or other document? Management might actually encourage the use of AI, not understanding its limitations. How likely is the employee to apply critical thinking to the AI generated sales report? Would he spend hours verifying, when the computer produced the documents in minutes? Work slop has the potential to do serious harm to companies that end up following hallucinated conclusions, because nobody bothered to verify how they got there.

Our first game of Covenant with 4 players took about 3 hours. Of that, 45 minutes were spent to set up the game and explain the rules. There are over 30 different types of tiles and tokens to set up, and the rulebook is 40 pages. This isn't a game that is quick to bring to the table, and it is also quite a table hog. I used several Gamegenic Token Silos to keep everything sorted, but as the game box is on the small side and very tightly packed, the token boxes then don't fit inside. You can keep everything in the box if you use the provided plastic bags, but that makes setup or storing take even more time.

Every action in Covenant has several steps, and some of these steps can then trigger additional steps as a reward. As a result, this is a game that requires a huge amount of concentration. We were constantly asking each other things like "did I take the victory points for this?" or "did I get the reward for this tradition?". Each move being so complicated also makes it rather hard to find the best move and optimize your actions towards victory. I mostly play casually, and Covenant isn't a game well suited to be played casually. While the rulebook isn't bad and has lot of examples and pictures, it is missing an index, and we had to look up rules questions all the time.

The strongest point of Covenant is its replayability: A great many things in Covenant are working via tiles, of which you have a larger stack, but only have something like 4 of them laid out. At the end of each turn, removed tiles get replaced. So from the start each game is different, and ever changing. Again that makes strategy difficult: Whether one strategy is better or worse than another might very much depend on which era objective tiles are laid out, and what personal objectives you drafted. My strategy in my first game, which was all about growing the strength of my dwarves and killing goblins, orcs, and trolls, ended up not being very successful, as I had neglected other things that brought more victory points. I placed only third out of four players, with the first two players being significantly ahead of me. Covenant rewards more hardcore gamers, who are able to calculate several complicated moves ahead.

That makes Covenant a good game if you play repeatedly the same game with the same people. After a game or two the game flow should improve, unless you have players suffering from analysis paralysis, for whom this would be the worst game possible. For my purposes, with board game nights in public spaces, where some of the players are more casual and groups change, Covenant is less ideal. I also dislike how Covenant is complicated due to so many different game mechanics, game pieces, and rules. I prefer games that create complexity more elegantly, with fewer moving parts, and shorter rules. Overall I am giving Covenant a 6 out of 10 rating.

First of all, what actually *is* a good investment? Of course there is no universal definition of "good" and "bad". Usually the best measure is return on investment, ROI, which is the annual profit generated by an investment divided by the cost of the investment. A typical "good" ROI for a capital investment project is between 10% and 20%, with tech industries usually in the upper range, while very old industries without much innovation might have lower, but safer, ROI. So, let's say that AI data centers, being new tech, would be a good investment at 15% ROI, so 15% of $5.2 trillion means we are looking for an annual profit of 780 billion dollars.

Where do investors believe this money is coming from? Mostly from replacing white collar workers by AI. According to the Bureau of Labor Statistics, there are 70.7 million workers in the US categorized as "Management, professional, and related occupations", which would be the ones most likely to be replaced by AI. Especially if we just talk investment in AI data centers, as replacing a plumber by AI obviously would need additional investment in robots, and this is rather unlikely to happen by 2030. So how much profit could an AI company make by offering services that replace such a white collar worker? Obviously companies would not pay more for such a service than the current cost of such a worker. That cost, as total labor cost, not just salary, is on average about $70 per hour, for 1,800 hours per year, or $126,000 per year. So let's say that a company is willing to pay $100k for an AI worker, and that the AI company has a great profit margin of 50%, so makes a $50k profit per human worker replaced by AI.

So we know that the AI industry needs to make $780 billion annual profit to be a good investment, and does so by making $50k per human worker replaced. Which means that the AI investment is based upon an assumption that over 15 million office workers will be replaced by AI. That is over 20% of the existing white collar employees. The US has currently about 7.4 million people unemployed, an unemployment rate of 4.3%. If AI is a good investment, we would triple the unemployment, which except for a short spike during the pandemic would be the highest rate of unemployment since the Great Depression. And that would have serious negative consequences to the rest of the economy.

So there are basically just those two alternatives: Either AI investment is a bubble, leading to a financial crisis, or AI is a good investment, leading to an economic crisis. The AI bubble bursting might be the preferable outcome, despite the serious effects it would have on the S&P 500, all sorts of index funds, and 401k pension plans. It also is the more probable outcome, because it is hard to imagine 20% of all white collar workers becoming unemployed without that causing serious social unrest and political fallout.

The obvious disadvantage of this plan was that I then had to get myself and my suitcase to the hotel by subway, but with my suitcase having wheels that was okay. In exchange I gained a strategic HQ for my foray into the fair, allowing me to store games there quickly, and to access stuff I didn't want to carry around with me all day.

I previously said that buying games at Spiel 2025 isn't great. The smaller stands of the developers sell only a few games, and are relatively accessible. But many games are sold at larger stands from the publishers, and those have queues where you can easily wait for 1 hour before you actually bought the game you wanted. However, Thursday morning I found that just walking around the Spiel was quite difficult as well, as it was very crowded. So I decided that if I was just shuffling in a crowd anyway, I might as well queue up for some games. As a result of that and the car parked nearby to store, I ended buying a whopping 10 full games, plus 2 smaller card games (Forest Shuffle Dartmoor and Innovation 4th edition). Here is what I bought:

1. Abroad
2. Altay
3. Covenant
4. Emberheart
5. Fromage
6. The Hobbit: There and Back Again
7. Kingdom Crossing
8. March of the Ants
9. Railroad Tiles
10. Recall

The one game I would have bought but couldn't was Feya's Swamp. On the BGG Spiel 2025 preview that game got over 1k thumbs up, but the publisher brought only 160 copies of the game to the fair, so it was sold out at 10:20 am on Thursday, with no additional copies for the other days reserved. The game comes from a smaller Chilean publisher, who hadn't predicted the hype that game would cause, and was thus woefully unprepared for the demand.

I had several great moments at the Spiel 2025: Buying a game I wanted and that won't be available in stores for months to come; getting information about a game I was interested in but not sure whether I should buy it; chatting with a developer for a game I had backed on Kickstarter; chance meetings with people I know. But the sad reality is that these great moments were diluted with a lot of boredom and hassle. For example I would have loved to get explanations on many more games, but typically each game was set up on just one or two tables; with 55,000 visitors per day, you can see how my slim my chances were to get an explanation for the more popular games. The games I bought, I mostly bought blind, based on previous research, as just seeing them set up on a table from a distance didn't really give me any more valuable information about them. Getting from one point to another at the Spiel meant shuffling slowly with the crowd. I stayed on Thursday until nearly closing time of 7 pm, and it was a bit better in the evening. But overall I had the impression that at the Spiel 2025 I couldn't see the games for the players.

I stayed in the hotel for another night, and had another day ticket for Friday, having planned to drive home in the afternoon. But in the end I decided to not use that ticket and drive home in the morning already. And I decided to not go to next year's Spiel in Essen, and instead just visit the much smaller Spiel Doch in Dortmund in April. While it is nice to get a stack of games early, and I now have enough for several months of board game nights, the cost per game this way is excessive, if you count in the cost for the hotel and drive. It was an experience, but not one I plan to repeat anytime soon.

Last weekend, 7 million Americans participated in "No Kings" protests around the country. Which means that 333 million Americans didn't participate in those protests. Trump's approval rating is holding pretty steady around 40% approval, 55% disapproval. That is slightly lower than other presidents, but not by much. So it struck me how some of the rhetoric from the left on the internet, calling Trump a fascist, is contradicted by an absence of effective opposition to Trump's rule. If Trump is a fascist, then what does that make hundreds of millions of Americans? Mitläufer? The "No Kings" protests had less than half the number of participants than the 1968 "Moratorium to End the War in Vietnam" demonstrations with 15 million participants, and America only had 200 million citizens back then.

The reality of history is that civil obedience is extremely common, and large scale civil disobedience extremely rare. Most of the Americans who disapprove of Trump have concluded that the solution to the problem will be the 2028 US presidential elections. I generally approve of the idea that sometimes the best solution to break the political power of a bad populist movement is to let them govern the country for 4 years, which shows everybody that the simplistic populist answers don't actually solve problems. If you compare Trumpism to other political movements named after their leaders, like Stalinism or Maoism, it is obvious that Trumpism doesn't have much of an underlying ideology, and is more about the peculiarities and vanities of Donald Trump himself. Don't tell JD Vance, but that sort of movement is unlikely to survive the departure or death of its leader. The probability of economic hardship, possibly a financial crisis, possibly a recession, between now and 2028 is rather high, and while not all of the problems are due to Donald Trump, Americans tend to blame their president even for the price of gas. Ronald Reagan's "Are You Better Off Than You Were 4 Years Ago?" is still the ruling principle of how America elects its presidents, and there is very little hope that the average American will be better off in 2028 than he was in 2024.

The question then basically becomes how much damage Donald Trump can do to the United States in 4 years. Fortunately, he isn't a very effective leader. There is a lot of bullying, extreme rhetoric, smoke and mirrors, but much less actual change. ICE raids maybe flashy and make headlines, but even if you consider their effect of discouraging illegal immigration, it is yet unclear whether the US will manage to deport enough people in 2025 to even just bring net illegal immigration to zero. Trump's idea to rule the country by executive orders has the obvious flaw of being easily reversed by the next president, or even being overturned by courts before that. When you consider that Republicans are still working on partially dismantling Obamacare 15 years later, and are unlikely to be able to completely abolish it, it becomes obvious how much more likely a president is to leave a lasting legacy when he works via the creation of laws rather than ruling by decree. The most significant lasting legacy of Donald Trump is likely to be a White House ballroom. That isn't much, from a historical perspective. In the end, Donald Trump will be less "L'État, c'est moi" and more "This too shall pass".

First of all, Malhya is a reflection of my typical Kickstarter experience. I backed the game in March 2022. At that time, the delivery was estimated to happen in July 2023. Actual delivery: October 2025, over two years late. While this is at the high end of delays, I had far more games that were a year or two late than I had games that arrived on time. On the positive side, I didn't crowdfund a single board game yet which wasn't delivered at some point in time.

Even if a game is on time, the length of the crowdfunding process for a board game is at least a year. Delays just add to the problem that if you buy a board game way in advance, your needs might have changed by the time it is delivered. In March 2022, I was still living in a different city, and my gaming groups were regular groups on weekends, playing long sessions of role-playing games or board games that resemble those. Thus Malhya is such a long campaign game, where every player plays a hero in a series of narrative adventures, with a regular group and long sessions. In 2025, the majority of my board game activity is pickup groups for board game nights in which we play one-off board games within a 3-hour time window. I still have two regular weekend groups for long campaign games, meeting between once and twice per month. Which means that it takes us often over a year to finish a single campaign game. Malhya is thus just joining a long queue, where previous decisions to back campaign games have caused a considerable backlog of those games.

Having said that, Malhya is looking like it could be a fun campaign game, of medium complexity. The box is huge, it weighs over 11 kg, and there are tons of miniatures, cards, tiles, and other components. Compared to other crowdfunded games I have backed, this one is from a smaller team, and there is considerable less hype and less YouTube content for it. Some would say that this is the better type of crowdfunded board game, with a very enthusiastic team trying to make the best game they could, and failings being mostly due to inexperience in project management. As opposed to larger companies that have mastered the art of good-looking crowdfunding campaigns for mediocre games. I am looking forward to playing this, I'm just not sure it will happen anytime soon.

As I now live just an hour and a half drive away from Essen, two years ago my wife and me tried to just go there on Thursday morning by car. It turns out that traffic around the fair area is extremely congested even hours before the fair. And while there are multiple parking garages, they are all full. We ended up parking illegally at the side of the road somewhere 10 minutes walk from the fair. So last year, we decided to take a hotel room from Wednesday to Thursday. By luck, we found a room somebody else had cancelled. Hotel rooms during the fair are extremely hard to find, and "surge pricing" triples the cost of a room to sometimes over $500 a night. This year, despite booking in January, I couldn't find a room anywhere near the fair, and had to take a room downtown. At least I booked it before their surge pricing algorithm kicked in, and got it still at the regular rate. So I'll have to take the subway from there to the fair this year, which slightly complicates things.

This year I booked the room for two nights, Wednesday to Friday. My experience from the last years was that getting into the fair itself on Thursday morning is hard. Many of the visitors queue before the entrance already well before the doors open, as they want to rush in early and get whatever hot game before it sells out. So this year I plan to turn up a bit after the opening time of 10 am. As in the previous years I always left the fair in the afternoon on Thursday, this year I plan to concentrate on Thursday afternoon and stay until closing time at 7 pm. I'll probably won't do that in just one foray, but rather go back to the hotel around noon, for lunch, a bit of rest, and maybe to store the games I bought in the morning.

That is, if I even manage to buy any games. The world's largest tabletop game fair is a surprisingly horrible place to buy a tabletop game. The place is just packed, and last year one stand selling a highly desired game on Thursday had a queue of over 1 hour to buy the game. Getting to the stand with the game you want is already a challenge. Getting to see the game is hard, getting 5 minutes of basic explanation is harder, and getting to play it is nearly impossible. Whether you can buy the game depends a lot on how popular the game is on the previews, and how many copies of the game the publisher actually managed to bring. Some games are sold out before noon on the first day. Last year, I only bought two games at the fair itself. Which was just as well, because the crowd in some places means that even just walking forward turns into more like a shuffle, and if everybody is carrying several bought board games, foot traffic gets even slower. Large Cajon backpacks are prohibited, but a lot of people have trolleys, which also make walking in a crowd difficult for everybody. I think I'll just bring the classic Ikea bag.

While getting a hot new game at the Spiel is nice, it isn't why I am there for. Ideally I want to see many different games, even games I am not yet aware of. I am mostly looking for games that can be played in under 3 hours, including set up and explanation, for my board game nights. That limits how complex the game can be. Then there are secondary considerations, about how much I like or dislike certain game mechanics, or how much the theme of the game appeals to me. If I like a game, and it isn't available to buy at the fair, I can still mail order it later. Seeing a game demo live is often giving me more information about a board game than a YouTube review video. I'll just have to overcome the logistical challenges and the crowd to get to that point.

Finally I got into the city and the real game began with me being introduced to the calendar system. This was new to me, as I haven't played any of the Persona games. I was told that I had 10 days until some calamity struck, so I had to finish a certain dungeon by that time. No problem, I thought, as the preparation before that, including doing city quests to reach Wisdom 2, which opened up the purification feature and a side dungeon quest, only took a few days. So I entered the main dungeon, and was starting to have fun. 8 hours later I had finished both the side dungeon and the main story dungeon before the time limit, when the game decided to slap me in the face and told me that I did it wrong.

You see, these dungeons have "rest" areas called Magla Hollows. But you can't actually rest there, only save the game and do some dialogue. They don't fill up your magic points, which are the main limiting resource in dungeon runs. However, then the game introduced me to the option of using the Magla Hollow or other save points (in the form of a cat) to teleport back to the entrance. From there I could exit back into the city, rest properly in a tavern, and then get back into the dungeon, teleporting right back to the Magla Hollow. Great! Basically exchange a calendar day for a mana refill. Should be fine, as long as I don't need more days than the time limit. Wrong! It turned out after the dungeon that you get any calendar days not used in the dungeon as free days, which you can use to increase your virtues, bonds with other characters, do quests, or grind. These free days are extremely valuable, especially at the start of the game. And I had basically wasted them, by using the outside rest option whenever my mana was a bit low, instead of wringing the maximum out of each magic point.

For me, good game design has failing forward options: If you aren't doing that well, the game should help you along in order to not fail at later, even more difficult challenges. Metaphor Refantazio reminds me of Valkyria Chronicles in as far as these games have failing backwards systems: They make the game easier for you if you played in a specific developer-intended "perfect" way, and punish you if you play casually, try out things, and are "inefficient" by very specific metrics. That is especially galling if you either need knowledge from previous games or guides, or get the information about what you should have done later in the game, and should basically replay parts of it "optimally" to get the best rewards.

So I read some guides and tips about Metaphor Refantazio, and was told something I otherwise would have learned only much later in the game: Each character has an ultimate archetype, but you can only play that one, if you ranked up very specific archetypes throughout the game. If you are again playing casually, and try out a wide variety of archetypes on each character for fun, you might get to a point where you simply don't have the time to still level the archetypes you need high enough to fulfill the conditions for the ultimate archetype. Unless you use a very specific exploit that allows you to farm infinite amounts of archetype xp. So, hey, I'm kind of happy that my Game Pass subscription ends in a week, and I won't be able to reach the end of Metaphor Refantazio anyway. Because if I have the choice of playing a RPG either for maximum fun while trying out a lot of things, or for maximum efficiency, I'll always choose the fun. Even if the game at some point would tell me that now I can't finish the game, due to all the accumulated penalties for not playing perfectly. I call that bad game design.

Apart from the failing backwards design, Metaphor Refantazio is fun enough in a very classical way. It does look a bit old-fashioned, compared to Clair Obscur: Expedition 33, which I just played recently. But the archetype system and the turn-based combat are all very entertaining. The story and world aren't half bad, although again a bit more classical: In my 12 hours I did fight a lot of wolves, goblins, skeletons, and a minotaur, but also a few more original monsters, weirdly called "humans". I am totally fine playing this game for another week, and then stopping somewhere in the middle. I'm just not willing to restart it, and play "optimally" in order to be able to reach some end game goal.

Of course they aren't the first ones. There are already lots of platforms offering "AI girlfriends", including paid-for NSFW options. But what struck me about the OpenAI announcement wasn't the shock that the internet could be used for porn, but the implied admission of failure: The valuation of all those AI companies is linked to a general idea that AI is going to make hundreds of billions of dollars by replacing a huge number of human jobs cheaper. Even if OpenAI "Erotica" would end up making as much money as OnlyFans, this is hardly in the same league of financial profits. Global AI investment in 2024 was $252 billion, which is about 4 times the size of the global adult entertainment market. You simply can't reach trillions in market capitalization with just porn.

For me this is another sign of the impending bursting of the AI bubble. I am seeing more and more comments in which analysts when asked about the risk of a bubble point out how just before a bubble the stock markets rise the most; sophisticated financial institutions have faster access to the stock market, and can bail out faster than retail customers, who would be left holding the bag when the market crashes.