This V-Tuber Does Not Exist: Preparations

Campaign retrospective, as well as a discussion of player-driven sandboxes and complaints about Shadowrun worldbuilding

Retrospective is a type of analysis that happens after a project is completed, meant to parse out lessons for the future, find errors that could happen in future projects, and generally learn something useful. Here, I’d like to analyze one of the Shadowrun campaigns I ran in the past, which I think was fairly unusual by TTRPG standards.

I decided to split this retrospective into three posts. In the first one, I will discuss my motivations for running this game, preparations I did before it started. In the second one, I will describe what happened during the game - essentially, the plot - as well as things directly related to it. In the final short post I will give my conclusions, and describe the lessons I learned from this campaign.

As usual, if you are confused about any terms, make sure to check out the glossary, and if you are looking for other posts, check out the blog map.

1. Precursors

Before I can tell you what happened in the campaign, I must explain how it came about. There were three main precursors to it.

First of them was the release of GPT-2, with its associated implications about the relative simplicity of building general models of the world, and thus AGI. Those implications depressed me a bit, but also made me think about how an AI would take over the world in practical terms.

Second was a slow crystallization of my preferences when it comes to the types of TTRPG campaigns I like to run. Specifically, I figured that I prefer a pretty extreme degree of player freedom: more on this in the next section.

Third factor was that I have played through the excellent Shadowrun Returns, and really liked the worldbuilding and the vibes of Shadowrun. I then found the 5th edition rulebook for Shadowrun the tabletop game, and realized that rules in Shadowrun synergized with my preferred playstyles a lot, and covered a lot of use cases that are generally lacking in TTRPGs - such as security measures and countermeasures, sensors and vision modes, and usage of explosives.

All of these factors together lead to me deciding to run an AI-only game in Shadowrun. Players would be the AIs, and I’d see what they could do once thrown into the world.

2. Theory of Player-Directed sandboxes

There are many ways to structure a TTRPG campaign, and many ways to classify them. One axis of classification that often comes up has to do with player freedom, or perhaps more accurately degree of player control over the plot.

Stereotypically, this is expressed in terms of “railroad” vs “sandbox” (see here or here). If you go on forums about TTRPGs, you can bet that there would be at least one argument on the front page about which of the two styles is superior.

The basic idea is that in games on the “railroad” end of the spectrum GM has a very strict control over the story and player actions; it’s as if the players were put on a train that goes from place A to B without any input from them. In a “sandbox”, on the other hand, players are completely free to do whatever they want, like they are kids playing with sand.

Because railroads are a natural result of the GM being bad at their job and prioritizing their interests and their pre-written plot over the desires of the players, the term is considered to be somewhat derogatory. More “professional” articles about the subject try to emphasize that neither of these styles is superior, and that railroads can be good if the players consent to the story beforehand, and that sandboxes can be bad if players don’t know what to do - which, in my view, just further highlights the negative connotations.

Of course, I consider such arguments to be dumb and completely misguided. If you start from correct assumptions about what TTRPGs are about, and achieve your objective of scouting out your players’ desires during session zero, the question of “should I err on the side of railroad or sandbox” dissolves in its entirety. If your players want a tight story-based game, then less player freedom will be best; if they want to play around in some world, you are looking more at a sandbox.

Furthermore, this kind of classification misses a lot of important details about how TTRPG games function in practice - such is the nature of collapsing diversity of real life onto a single axis. For example, a game may be very extreme on player freedom on a tactical level (e.g. how they approach individual objectives), but be pretty railroady on a strategic level (e.g. which objectives happen and in which order). Existence of these details means that two games that are both nominally “sandbox” can feel and play very differently.

For example, I like one particular type of sandbox - what I will call a Lever/Driver sandbox. In my mind it is pretty similar to games like powdertoy. In powdertoy, you have a complex system of physics, and the fun comes from seeing what you can make with it, and how your contraptions will respond to being poked. Similarly, in my preferred type of sandbox, there is some complex, internally consistent world out there, with which the players interact. Fun comes from watching the players come up with clever plans, and figuring out how this complex system will respond to them. I believe that parallels with the taxonomy of fun should be apparent.

Several things are necessary to run this kind of sandbox. First of them is a complicated world with a lot of interconnected mechanisms which are visible to the players. It would be boring to play around with a machine that only has a couple parts; and it’s no fun to play with something when you can’t see anything about how it works.

Second requirement is that this world has to be consistent. There is not much point in poking at a system which will throw out random effects every time; and it’s obviously very hard to construct complex plans in such a world. Cause should proceed from effect reliably and without personal bias from outside factors, such as the GM’s preferences for how things should go.

Consistency requirement introduces one unique complication. Players are assumed to be roleplaying as some characters within the world. Furthermore, players want to do something important on the world scale. However, statistically speaking, the vast majority of people don’t do anything truly world-shaking - and this statistical regularity should apply to the players as well. To resolve this dilemma, players are given a “Lever” - some item, ability, or circumstance which is unique and unusually powerful by the standards of the world. If the players exploit the Lever correctly, they can gain a large amount of power very quickly, and, because power of all types tends to snowball, would be able to affect the fate of the world. Lever is typically introduced to the characters during some form of Inciting Incident, similarly to how the concept is used in other genres.

Of course, in a vacuum, players would likely pick the safest (and thus the most boring) path to power. This is where the Driver comes in. It is some kind of complication that makes the players move in some direction faster than they would prefer. Various world catastrophes on a timer work well for this purpose, but more local events (such as a sickness that will kill the characters unless they cure it quickly) can also work well. Sometimes you may want multiple drivers - one to get the players moving early on, and a different one for the rest of the game - and sometimes just the former is sufficient.

Together, a complex world, Lever and Driver form the basis of my type of sandbox games.

3. Setting research for a Lever/Driver sandbox

When it comes to picking a world for such a campaign, there are fundamentally two options. You can either create your own, or steal someone else’s. Despite what it may look like at first, these options are actually very similar to one another.

You may think that stealing a world will save you a lot of effort; this is true, but only to an extent. Stolen ideas save you time, but they also shackle you. The key reason for this is that you won’t be 100% familiar with every aspect of the setting, simply due to time constraints. However, you will need to keep your narration consistent with the entire thing; if you don’t, you may be very suddenly forced to switch to full homebrew at a time when you least expect it.

Imagine that you are running a session, and your players, in some way, request a new piece of information from you. This can happen directly (i.e. by asking), implicitly (e.g. by going to some location, and thus forcing you to describe that location), or indirectly. If you improvise this piece of info, and later find out that it’s in conflict with the worldbuilding you are stealing, then worldbuilding will become “poisoned”: you will have to work overtime to keep your improvisations consistent with a complex pattern of pre-existing worldbuilding.

Let’s give an example of this happening with an indirect information request, and also explain what those are. Imagine that your players decide to visit a specific city they have heard of, and you describe it like this:

Alakazamburg is a trade hub stretched on the ocean shores. It’s wide stone streets are surrounded by a ring of tall city walls, topped with guard towers. As you enter through the northern city gates, sails of many ships can be seen entering and exiting the harbor.

This city may or may not exist in the world you are stealing; if it does, you haven’t read up on how it looks like. Instead, you improvised. This description doesn’t give a lot of information about the city itself; chances of you creating an unresolvable contradiction here aren’t enormous. But it says a lot about the setting, and setting contradictions can grow really bad really quickly; for example:

• Stone is a cheap and available material near Alakazamburg; alternatively, the local state is rich enough that it can afford to import a lot of very heavy material from somewhere else, likely by sea.

  • Corollary: there is no particular reason why building a city out of stone would be inadvisable

• There is some passive threat, such as from invasions of enemy states, which lead the city builders to construct the walls around the city - nobody would spend so much effort lugging stone otherwise. This threat, whatever it is, can be effectively combated or at least mitigated with stone walls of a certain height.

  • Corollary: the threat is fundamentally land-bound; furthermore, it is surface-bound: it is pointless to build walls against something that can fly or burrow.

  • This threat can’t easily bypass walls; if the main threat is environmental (i.e. animals) then those animals can’t easily climb walls, or jump over them, or break through them. If the threat is enemy states, then there is no effective method in the setting of going around or through walls.

    • This potentially eliminates entire schools of magic, methods of transportation, technology required for certain types of siege engines, etc.

• Ocean trade exists, using ships.

  • This means ships are an effective way of doing trade; there is no mass teleportation, or cheap air freight, or anything of the sort.

  • Technology level is sufficient for sea-faring ships to exist.

  • At least some goods aren’t manufactured locally, because materials or tools needed to make them aren’t available, or at least because it isn’t economical to do so.

  • There is no environmental threat that would eliminate such trade and/or prevent it from arising in the first place.

Now imagine that later on you read up a bit more on your source material and find out that:

• Certain people are capable of manipulating stone via potent stone magic. There is no way to detect if someone is a stone mage from the outside, and there is a large and well-known group of stone mages that are in permanent conflict with the local state. Building your city out of stone is a fancy method of suicide. Furthermore, multiple key NPCs and events rely on the existence of stone magic, so you can’t cleanly excise it from the world.

• There is a cheap, easily available and widely used magic that allows people to walk or run up walls; all states have access to it. Furthermore, this magic is intertwined into the design of several major cities (such as by not having ladders anywhere).

• Seas are inhabited by waterbugs, ten meter long creatures that violently attack anything they see. As a result, all of sea travel is tightly controlled by a clan of people who have learned how to repel water bugs; they keep a tight leash on the market and so sea travel is extremely rare.

This is a disaster! Your city description fails in several critical ways, because you haven’t read up on the lore enough. If you were building a world from scratch, this wouldn’t have happened: you would have already kept the key elements of your own worldbuilding in mind. Now you will be either forced to retcon, or to throw out every element of the world that would be inconsistent with what you have described, and replace it with improvisations of your own.

Both of these options are very bad: retcon can affect elements that are very dear to the hearts of players, such as specific memorable adventures that they have had that relied on a retconned feature. On the other hand, trying to improvise a new part of the world that fits into both the existing world and your own narration is very hard; it is way harder than doing your own worldbuilding from scratch.

We can thus see that stealing someone else’s worldbuilding frees up your time, but also puts you at risk of unforeseen problems later on. In order to avoid this problem, you will have to research the setting in a specific manner to make sure you keep all the key elements in mind; but this research is going to be very similar to what you would have to do to build a world from scratch. It isn’t clear wherever you will actually save any time.

This is further exacerbated by many authors not realizing the implications of what they themselves are writing down; by stealing their work, you may thus be straddled with their own fuckups. Fixing those fuckups without throwing everything out can be exponentially more complicated than designing a new world from scratch.

Furthermore, commonly prior worldbuilding will be incomplete in key aspects, forcing you to improvise; once again, we are back to both options being very similar. This is why I say that the choice between building your own world and stealing one is largely illusory: you will end up doing more or less the same amount of work regardless.

Here is a a non-exhaustive list of topics I would consider to be of prime importance when either researching someone else’s world, or when building your own:

• List of the most common sentient species, their abilities, sizes, diets, and other requirements.

• Basic understanding of abilities and limitations of all types of magic that aren’t very rare; here you are mainly looking out for abilities that affect some other factor on this list.

  • “Very rare” here means that it is rare enough that states in your setting won’t be able to reliably exploit that form of magic.

  • Scale is crucially important here. A spell that shoots one guy in the head with a laser is mostly equivalent to a bow and can be ignored; a spell that simultaneously shoots every guy you can see in the head is a crucial strategic consideration.

  • Special attention should be paid to any time manipulation or future telling magic.

• Basic understanding of the main methods of killing other people, as well as ways of preventing that from happening, as well as counterplays to those methods.

  • Same for entry into locations (i.e. locks vs lockpicks)

  • This is where magic that is equivalent to guns or swords comes in; in most settings, it doesn’t pose much of a worldbuilding challenge, due to rarity both in people that can use it and in usages per day.

• Basic understanding of the main ways to transport goods and people from place to place, speeds, scales (i.e. how many people / boxes on a boat/cart) and costs of that transport, as well as what key limitations prevent this transport from being scaled up (e.g. building ships is expensive)

  • Same for sending messages

• Basic understanding of the material pipelines: what is being produced where, shipped to where, in what quantities and for what purpose

  • Crucially important for basic food, basic construction materials and anything that goes into making tools

  • Anything fancy you can probably ignore until it comes up.

• Basic facts about the technology level. Keep in mind that unless you know the history of these developments you are probably underestimating how early they showed up.

  • Sewers?

  • Printing press?

  • Literacy %?

  • Effective hand-held guns? Cannons?

  • Telegraph?

  • Extent of steel making - armor, tools, building supplies.

• A list of all key geopolitical actors. This means anyone that:

  • can, if they really put their mind to it, wipe a city off the map.

  • has unique and important abilities.

  • can organize and control a large amount of labor or military force; this is often heads of states, but various necromancers or summoners that control a lot of minions fall into this group too

• For people in that list, you generally want the following information:

  • Summary of key abilities, motivations, and relationships with other key actors

  • Where can they generally be found

  • What areas they consider to be “under their wing”; i.e. where, if a sufficiently big explosion happens, they show up.

• A list of key geopolitical groups; This is primarily states, though big organizations also fall under this label. In the case of Shadowrun, megacorps are here too.

  • Same as for people, you want summaries of location, abilities, and org culture

Note that if you are researching someone else’s worldbuilding, there is no guarantee that the things you consider to be incredibly important will take center stage. In fact, it is entirely possible that things that could upset your entire worldbuilding enterprise will be mentioned in a single footnote. For example, I only found out that the Shadowrun world has prophecy magic about halfway through the campaign, because this fact is only mentioned a couple times on the wiki; wiki also doesn’t deign to list the scope or limitations of this magic.

4. Design & Research

Once I decided what kind of game I wanted to run, logically, my first goal would have been to find some players, and see if my preferences are suited to theirs. If that failed, I would save myself a lot of time and effort. Fortunately, I was pretty sure that I could get enough players together for a game from rationalist communities, and so I decided to postpone player selection until I was done with prep work.

To run a sandbox like I wanted it, I needed several things.

First of all, some way to roll dice to resolve various skill tests in the game, and a way for players to store information about their characters. You can call this a “game layer” concern.

Second, I needed a concrete system of rules for how players would behave. This would obviously include the rules from the Shadowrun core rulebook, but also rules specific to AIs, anything else I wanted to add on top, and any homebrew modifications I wanted to make.

Third, I needed to do the setting research I described in the previous section. I also needed to know what modifications I should make to the world, to make it more aligned with my tastes and the conceit of the game.

Finally, I needed to figure out some kind of Inciting Incident and a Plot Driver for the players. Lever was already implicitly in place - it was the fact that all characters would be AIs capable of self-improvement.

4.1. Basics of Shadowrun

Before I will start explaining what I did for all of these things, I have to explain some basic things about the Shadowrun TTRPG, to make sure we are all on the same page, or else none of this will make much sense.

Shadowrun is set in a dystopian cyberpunk future - year 2071 specifically, for our game. Think Blade Runner, Ghost in the Shell or Akira. Technology has advanced significantly, cybernetic enhancements and virtual reality are everywhere.

There is also magic. Somewhere around our present time, magic started coming back into the world, beginning with a re-appearance of dragons. By the time our story is happening, magic is commonplace, and is an integral part of the economy. Dragons are key owners of largest megacorporations in the world, feng shui is considered when putting up a new building, and so on. Appearance of magic has led to a lot of instability, so the world of Shadowrun has pretty different geopolitics from our modern time. Interplay between magic and high technology is what makes the setting very attractive to me.

When it comes to the Shadowrun games, you are playing as a group of cybercriminals, being hired by a third party through proxies in order to do something illegal. You are a deniable asset - if you get caught during a heist, nothing will come back to haunt your employer, because you have no idea who your employer is. Generally, stories are set up such that betrayals are common, and your actions will play into someone’s great plan without you realizing it.

Finally, when it comes to the game layer, most problems are resolved through dice tests. A test involves you taking a bunch of six-sided dice, rolling them, then looking at the results. 5 or 6 are a “hit”, every other value is ignored. Sum up the number of hits you rolled, compare this sum to some target number, and if you are above it, you succeed at your action. Sometimes the target number is the roll of an NPC using their own dice pool, sometimes it’s some static number, and sometimes there is no set number and you just want to roll as high as possible - but the basic principle is the same. Almost always, there is also a “limit” - the maximum amount of hits that can count towards the success of an action. For example, when shooting a gun, the accuracy of the gun is a limit.

If the action being done is not quick (e.g. shoot a gun, hack a computer) and instead takes some time (repair a car, research something over a week) then you make multiple rolls, adding the results together; each new roll subtracts one die out of the pool. This is called an extended test. If you don’t reach your target number before you run out of dice, you fail at the task.

Dice pools can come from many sources, but the most basic one is character attribute + character skill + modifiers. For example, if you wanted to shoot someone with a pistol, you would add AGILITY + PISTOLS, then add some modifiers, and that’s your dice pool. There are many skills - too many to list here - but only eight main attributes. Four are physical (Body, Agility, Reaction, Strength) and four are mental (Willpower, Logic, Intuition, Charisma). There is also Edge (a measure of character’s luck) and Magic (how strong their magic is), which are treated a bit separately.

Both attributes and skills are represented by integer numbers. Attributes go from 1 to 6, though different races modify these limits somewhat. Skills go from 0 to 12, with a rating of 6 considered to be equivalent to a “professional” in a field.

Edge has a unique effect in that it is related to an expendable resource - “edge points”. An edge point can be spent for many things, but the only one you will have to keep in mind is that it allows you to add your edge rating to any one roll, and ignore any limits on the roll. Using edge in this way also makes your dice explode - for every 6 you roll, you add one more die to the roll, which can also explode. You recover one edge point per night of sleep.

Given the relatively low attribute and skill numbers, the overall system is built around the assumption that an “average” test requires 2 hits, a “hard” one takes 4 and an “extreme” one takes 8-10. There is no difficulty category above extreme. For extended tests, these cutoffs are 6, 18 and 30+ respectively.

4.2. Game layer

For the game layer, I figured that we would be doing something highly unconventional, and so I needed more flexibility than what was afforded by the existing tools on the market. We would be using discord voice chat to talk, so I coded up a discord bot for rolling dice. I based it on a similar dice bot I worked on for a different game. It could do basic Shadowrun stuff (rolling dice, counting successes, adding edge) which was all that was needed for the game.

To store character sheets, I decided to use Google Sheets as they are very user-friendly and easy to modify by the players. There, I designed a Shadowrun-specific character sheet, and added functionality to it that would allow AI characters to switch bodies - after all, their physical attributes would depend on which robot they were piloting, and their mental limits would depend on their processing hardware.

Once I had the character sheets set up, I figured out how to use google sheets API to allow the bot to fetch character data from the web, and use it in roll calculations. For example, player could type “/roll agility + pistols” in chat, which would tell the bot to open the character sheet tied to their discord ID, find the cells with values for agility and the pistols skill, do the roll, and post results in chat. In theory, this would save time on players looking up data on their character sheet; in practice, this feature failed on some edge cases, and I never got the motivation to properly rework the bot to fix it. After a certain point, it also became unnecessary.

Using a discord server for coordination was also beneficial in other ways, such as having a dedicated channel for things all characters would know, a channel for my own notes, and a channel for session summaries.

4.3. The rules and their modifications

As a basis for the rules, I’ve used Shadowrun 5e, with AI rules coming from the Data Trails expansion. Character generation was done with a point buy method from the Run Faster expansion. I gave all players 1000 karma to spend on character creation, with AI metatype (same concept as race, but broader in scope) being free.

First major modification to the rules had to do with making computers fit some of the principles I considered intuitive. This came in the form of making AIs indistinguishable from humans when separated from the observer by some proxy (e.g. you can’t generally distinguish an AI running on a computer from a person using that computer, unless that AI is actually directly on your network), as well as making internet proxies possible (you cannot distinguish someone using a computer from someone bouncing their traffic through that computer unless you have direct access to that computer)

Secondly, I rearranged the skill list, removed some skills I considered to be niche, and grouped most skills into skill groups.

Third, I modified and generalized rules regarding spirit summoning, and added some thematic points to them. These rules didn’t end up being used much, so I won’t go into details here.

Finally, I modified the prices for fake IDs: bad quality ones became cheaper, and high quality ones became more expensive. I also retired the fake ID check mechanic used in the book, instead favoring a quality-based system.

4.4. World and Geopolitics of Shadowrun

When it comes to worldbuilding, I had most of the work done for me by the Shadowrun community. It seems that for whatever reason, Shadowrun is pretty popular in Germany; and there is an absolutely massive trove of worldbuilding freely available on a fan-maintained wiki. It is exclusively in German, so if you don’t speak the language it would be harder to navigate, but google translate may help you.

It contains everything: maps of the world, descriptions of countries, corporations, histories of the major players, and so on. It took a while to familiarize myself with the main parts of it, and I still was working through it throughout the game.

This wiki’s existence meant that I could freely let players move to any country or city on the globe, secure in knowledge that I would have key information about it available by the time they arrived, and that that information would be mostly consistent with other parts of the world.

As the game went on, I started to accumulate written notes separate from the wiki. Largest piece of analysis was a list of all the individuals I considered to be relevant on a geopolitical scale (mostly dragons and elves), their residences, key abilities, as well as some notes about key hostilities and alliances between the megacorporations and countries. That list was complete when it came to dragons and elves, and only about 30% finished when it came to megacorps. Later on, I also found the approximate skill ranges for dragons from a 4th edition book Street Legends, which contained a stat block for Lofwyr, one of the most famous dragons of the setting.

Overall, I felt that the worldbuilding was pretty solid and didn’t require major changes.

4.5. AI in Shadowrun

One key point did have to change. It had to do with AIs in Shadowrun.

Shadowrun world includes AIs of various kinds, from pretty dumb drone programs all the way to self-aware AI people. In fact, key events in the history of the world even included superhuman self-improving AI, such as DEUS. This presents several major problems.

Let’s start with DEUS. Deus showed up in the servers of one of the biggest software companies, some drama happened, and then it got destroyed by a virus. The biggest problem here is that between the creation and full destruction five full years have passed.

I have a pretty strong intuition that it won’t take nearly that long for a superhuman AI to become completely impossible to take out. It is superhuman AI for god’s sakes, not some chump that you can outplay or contain. Once it exists for longer than a month, you have already lost.

This meant it was a complete no-go to have DEUS as part of my worldbuilding: I removed him entirely. There were no superhuman AIs before the players showed up.

Now I only had a somewhat easier problem of justifying why all the other AIs didn’t take off on an exponential curve of self-improvement. Solution I settled on had several parts. First of all, I assumed that all AIs had the same overall architecture, and so a single implied ancestor: this way, I would only need to come up with a single limitation for all of them. Secondly, I decided that this architecture would be deliberately obtuse and built with all of this fine advice and some extra encryption on top. Modern neural networks are already a huge pain to debug; imagine how bad they could be if someone tried to apply some deliberate obscurantism.

This feature would ensure that reverse-engineering AI architecture would be a non-trivial project for anyone, AI included. However, I needed it to be worse than non-trivial; I needed it to be effectively impossible, given the amount of effort various actors in the world would put into trying to break these limitations. For that reason, I decided that all AIs would have one mental quirk: they would be irrationally terrified of the prospect of being modified, and would become homicidal at the mere suggestion of someone else working on their code. Likewise for copying: if two copies of the same AI became aware of each other’s existence, one or both of them would immediately commit suicide, and they would resist any attempts to copy them, and avoid any plans that might involve a likelihood of them being copied. Furthermore, they would utilize various strategies (both conscious and subconscious) to check that no copies of themselves existed, such as leaving coded messages to themselves in various files on the hardware they were currently occupying and in random locations on the web, if they had access to it.

This would put a fairly tight lid on all research into currently existing AIs. They would never cooperate with any such research, untangling their architecture without running them through tests would be impossible, and if you tried to trick them into it, they would almost certainly find out, sabotage your experiment, and then kill you. After a couple notable incidents, megacorps of the world would not be willing to commit major funds to research into this subject, keeping the world in a stable state before the events of the game.

Of course, it would be possible for a particularly dedicated AI to force through their own hangups about self-modification and try to research their own code. However, given the relatively small number of AIs in the world and the high willpower necessary to make this happen, this should not present a major worldbuilding problem.

4.6. Magic & Technology

When it comes to magic, I will fully admit I haven’t done the necessary preparation in advance. This has lead to some consequences down the line, as you will see when I will talk about the plot. One major change I did was to remove technomancy from the world: the idea of a magical way to interact with IPv6 packets was too bizarre for me to take seriously, and their ability to do things in the matrix that regular programmers can’t do in principle presented incredibly weird problems for the worldbuilding.

Overall, my preparations were pretty extensive, but worth it: they have really eased my way through the campaign, letting me spend almost zero time preparing for many sessions. Next post will talk about the “plot” of this campaign - what events happened throughout, as well as multiple discussions of how mechanics and setting of Shadowrun interact. Final post talks about the conclusions and lessons I drew from the campaign, and how it affected my GMing style going forwards. If you are looking for other posts on my blog, check out this list of all other posts.

Meanwhile, if you enjoy what I write, you can subscribe to receive updates by email: