Oct 22, 2019

AI War 2 - x-4000 (Chris Park)

Chris here! By any sane metric I can think of, AI War 2 shouldn't exist. And yet it's more than I ever imagined it could be.

When we first set out to make this game three years ago, it was far less ambitious. Even that was going to be really hard. Somehow, in the process of falling down the stairs over and over again during this period, we wound up with a game that seems to be superior to the original.

This game shouldn't exist, but it does, and I'm both proud and stunned.

The Secret? Community

The intro to this post runs the risk of sounding like I'm tooting my own horn, but it's not actually about me. I'm not capable of making -- even just designing -- this game on my own. I don't think anyone is, really.

The fact that this game exists isn't a testament to me having some brilliant insight or a singular vision that I doggedly pursued. The reviews of the game are lovely, but give me entirely too much personal credit.

The state of this game is thanks to dozens of people critically thinking about this game -- what this one and the original means to them and to others -- and then a process of relentless, continuous, arduous iteration and improvement.

MVP Award: BadgerBadger

This section is long, and in some ways tangential, but if you read it you'll understand why I've put it here so prominently. You have this guy to thank as much as me or Keith for this game existing.

Badger has been involved since the kickstarter, with questions and comments and key insights. For a lot of the first year, he was responsible for something like 80% of the bug reports and feature ideas on our idea tracker. When a lot of other people were just bouncing off the game and waiting around for Keith LaMothe and I to figure things out on our own -- understandable, really -- Badger was there providing really key insight and ideas.

But that was barely the start for him. After a while I was essentially like "so, do you just want source code access, given how much you're doing here?" Because he had started doing some mods -- nanocaust and macrophage, at the time, IIRC -- and it was clear he would be less hand-tied if he had more access.

What happened next was essentially us getting a developer -- volunteer, no less -- who contributed as much to the design of the game as I did, in my opinion. Not only did he single-handedly conceive of and implement the nanocaust and macrophage, but he also did the dark spire and marauder impelementations, among many, many other things.

Some of the most brilliant and devious things that the AI has in this game compared to the first one? Badger. Some of your favorite UI detail screens, like metal flows? Badger.

Not to mention all the bugfixes, balance tweaks, and... just oodles more. This game wouldn't exist in anything like the state it does now without Badger. Any credit for my "singular vision" on the game is doing him a major disservice, but he's a quiet sort of guy when it comes to taking credit, so I wanted to take this chance to call him out in particular.

Growing Volunteer Developer Corps

So, Badger is not remotely the only person I need to call out as being absolutely indispensable.

RocketAssistedPuffin has also been involved heavily for the last year plus, and has taken over huge numbers of things that I never would have had time for. After I implemented the new tech system one way, he's the one who figured out how to make it substantially more balanced. Most of the voluminous "balance change" sections on the release notes are from him working with other players or just reasoning things out himself.

Puffin has also had a ton of ideas on how to make things better in all sorts of sections of the game, and there was a period of about three months late last year where he and Badger were basically doing ALL the development on the game and I just pushed out releases of what they were doing. I was going through a really painful divorce and had a ton of anxiety and couldn't face work, and these two kept things alive and improving.

But it never stopped there. Those new tutorials you like? Puffin. I wrote the bulk of the "How To Play" in-game wiki sections, but the most basic and understandable ones for new players were... again, Puffin. I'm excellent at writing encyclopedic entries that fill you in on huge numbers of details, but he's the one who distilled "what's the most central stuff you need to know, as briefly as possible" so that people can actually get into the game in any reasonable timeframe. Compare his work here to the tutorials I did in the original game, and it's night and day.

And I'm still selling Puffin short, frankly, because he's done so many things over such a long period that I can't remember it all now.

More recently, we've had folks like WeaponMaster and Asteroid joining in and adding lots of bugfixes and quality of life improvements that I never would have had the time to do myself. Things like hovering over galaxy map links to see information on them were Asteroid. Endless tricky bugfixes were WeaponMaster. I'm selling them both short, but the release notes are filled with things that they either implemented or suggested or both.

And it doesn't stop there. Quinn stepped in an made a bunch of additions. Keith laid the original groundwork for the entire game simulation and multithreading (he was the main programmer and designer for the first year and a half, and actually on staff during that time).

And there were so, so many others. And more each month!

Volunteers Beat Modders, I Think

I'm pretty free with the source code access, because I'd rather have a consolidated community of people helping rather than a bunch of mods that you have to hunt down and find.

So a lot of the folks that have turned into volunteers are what would have been modders on most other games. They would have made their own thing that you had to install and then wonder about the cross-compatibility of.

This game does have a ton of moddability, and for anyone who wants to "just" be a modder, that's absolutely fine with me. But for a lot of the mods that are getting the most love, I'm happy to share source code access with those folks so that they're in no way hobbled, and so that their work can go out as additional content that every player can find via in-game options without having to hunt through Steam Workshop or whatever else.

It's an unorthodox approach, but a lot more team-oriented and lets us do quality control on each others' stuff, "mods" included, which is a big win. If someone wants to steal the source code for this game, they can just decompile it like any other game for the most part. I'd rather put my trust in people and see things flourish rather than retain a stranglehold out of fear or pride.

What Did I Actually Do, Then?

All of this help from others let me focus on some of the really tricky architectural and design problems, which led to things like us even being able to HAVE a simulation of this size, and to have it perform as smoothly as it does.

I got to build lots of mechanics that other people then actually turned into specific units. It also gave me time to focus on some really nagging problems that just made the early versions of the game... unpleasant.

If I hadn't had the time to think and talk to people about all those things, we never would have seen all the game evolve this way; I would have been mired in content development and other items just to get the basics out for the game.

The original design for this was something that Keith and I put together as a pair, but it only worked out so well. It was a good foundation, but needed... a lot of help. We both pushed that forward a lot, until the money situation got to the point where he (and all the other staff, eventually) had to step away, and I carried on "alone" (but with all those volunteers).

There came a couple of major turning points where I was reflecting on why I was so unhappy with this game as it existed, and listening to the various gripes that playtesters had, and then I was able to spend a month or three implementing something drastically new.

Fleets are the most notable of those, and they were initially met with a lot of mixed feelings and distaste because only part of my idea was there on the first public launch of those. Only in the last month or so has that feature completely come into its own, and that also had a lot to do with continuous feedback from people in early access telling me what they needed and what they did and didn't like.

We also had a number of points during development where we just couldn't escape certain performance problems, because there were suddenly battles that were an order of magnitude larger than the first game (which was itself the largest strategy game simulation of individual units that I'm aware of on the market until this sequel). So I got to focus on a whole bunch of crazy improvements and data structure inventions and even GPU shader tricks in order to make all this stuff work.

Without the rest of the community helping, there's no way I could have had time to work on all that sort of thing, even in three years of development. A game of this scope shouldn't run this well -- it shouldn't be possible -- but it is because I was given the gift of time by so many others.

A Decade In The Making

It has been 10 years to the day since the first AI War came to Steam, and it's been 3 years of developing this sequel.

We didn't do any work on any AI War games from late 2014 through late 2016, but the rest of that time has been spent at least partly working on the original game or this sequel.

From version 5.0 of the original game through version 8.0, Keith was pretty much the sole developer on that while I focused on other things. He built out a ton of creative and clever things that made a return in this game, and also pushed the concept of what the AI could be -- adding in some traditional decision-tree style logic in places in addition to the more decentralized-style AI that I had come up with back in 2009. That one that originally made waves on slashdot and reddit and hackernews and so on.

I've worked as the producer and design lead on this sequel, among my many other roles, and so the fact that there seems to be a "singular vision" is hopefully a sign that I did a good job in that role. But the degree to which this is a product of dozens of people's work, over an extremely long period of time, really can't be understated.

That's what I meant all the way back at the start. This sort of thing shouldn't have happened. It's just so... unlikely. A ton of people came together over a decade and helped make something unlike anything else on the market.

That's before even getting into other major (former) staff contributors like Daniette "Blue" Shinkle doing the vast majority of the art and coming up with the way-prettier style of ship that evolved AFTER the kickstarter, the awesome score by Pablo Vega, and 25 voice actors who did a fantastic job as various humans and the AI.

And good grief, I'd be remiss not to mention Erik and Craig and all the other folks at Indie Bros, who helped manage so many aspects of this game, as well as often doing work like helping clean up voice lines, etc.

A Few Common Questions

If you're interested in what is coming in the very short term, there's a post for that.

Similar if you want to know what the plan is for multiplayer.

For kickstarter backers (or anyone else who is curious), there's an FAQ as well as roadmap of stuff for the next few quarters relating to kickstarter stretch goals.

And I just have to once-again plug this awesome After Action Report by zeusalmighty.

My Deepest Thanks

I never wanted to make this game, because I didn't think I could. The original AI War seemed to be the high water mark of my career, and I spent a lot of time trying to make peace with that. But when the market shifted in 2015 and 2016 and finances started getting tight, it became clear that returning to the game that started it all was what made the most sense.

Thanks to all of the kickstarter backers for believing that we even COULD build this game. Keith and I felt like we could do something that would make you happy, but probably not something that would top the original. It took two extra years of development and an enormous village of people to make THAT a reality. So thanks to everyone for their patience and support during that time.

I also want to say a big thanks to everyone for their understanding during my divorce, which happened shortly after entering Early Access for this game. That made everything so much harder, and took me out of commission for a full three months or so where I just couldn't work much. I had to learn how to be me again, and come to terms with being a dad with shared custody rather than a full-time father, and all of that was incredibly hard.

But the good news is that, as has happened with this game itself, a lot of things in my personal life have turned out unexpectedly, improbably well in the last year. After deciding to date again (after 18 years off the market, wow), I wound up meeting the woman who is now my fiancée surprisingly quickly (all things considered). Kara and her daughter have made my entire world so much richer than I realized it could be, and my son finally has the sister he's wanted for so much of his life.

I feel incredibly fortunate, and a lot of my ability to get back to work and not crumble under the weight of anxiety and expectations for this game were thanks to Kara's support and presence. The reality of her life as a doctor and surgeon also helps to kind of put my own work into perspective, sometimes, in the best way.

However this turns out financially, and despite my anxieties about my future as a game developer, I'm incredibly proud of what we've all created together, and I feel surrounded by all the right people both at work and outside of work. This has been the hardest three year period of my life, but the end result has all been worth it.

Thanks to everyone, and I hope you enjoy the game -- both what it is now, and what's to come.

Very Best,
Chris

Oct 22, 2019

Command: Modern Operations - The new simulation & editor features

Command: Modern Operations - SlithFox

Command: Modern Operations is coming soon. In this piece, let us take a closer look at some of the major new features integrated into the new editor.

Turbo-charged for your pleasure

One of the most immediately noticeable features, especially when running large or elaborate scenarios, is the jump in performance. Command now offers both greater absolute speed in simulation execution (ie. reduced “pulse time” as displayed when the “Show Diagnostics” option is enabled) as well as significantly increased CPU utilization, as a result of improved parallelism. (As an example of the latter, on a recent trial run of “Northern Fury #9: Hold the line”, CPU load was observed fluctuating between 65%-75% on a Core i7-6700 system. This may vary across different systems, of course). Maximum practical unit count has also increased thanks to better memory management, with the result that even more, massive scenarios are now feasible.

Until the COWs come home: Comms disruption, aircraft damage & new weapons

As we previously mentioned, it was decided to merge the previously DLC-locked features of “Chains Of War” to the core simulation available with Command. This allows both scenario authors and players to make extensive use of these features without any licensing concerns. We’ve covered these features extensively before, and we included cargo, amphibious and airdrop operations on our previous article on ground warfare, so here is a short summary of the others:

- Comms disruption explores the different & numerous ways that a unit can be isolated from its parent side’s communications grid, and the repercussions of “going offline” (spoiler: none of them are pleasant). This feature has been used very successfully on COW, Shifting Sands and The Silent Service, and now it can be used on any scenario. One of the post-launch improvements of this feature is that when a unit re-joins the comms grid, the UI provides much more finegrained feedback on what type of information was updated for a contact (if any), and skips mentioning contact merges that produce no new info. Ready to feel a cold sweat as your units abruptly go out of visibility and out of your control?

- Detailed aircraft damage is arguably the most popular feature of the bunch, enabling a much more finegrained level of damage & attrition on aircraft than classic CMANO’s “kill on first hit” paradigm. Since its original introduction, this feature has been significantly enriched with feedback from both commercial & professional users.

- Advanced weapon types: High-energy lasers, railguns & HVPs, and tactical EMPs are now available at your disposal, to mindlessly include in the next Transformers flick use in all scenarios addressing the challenges of warfare in the foreseeable future.

Alone and unafraid: Realistic submarine comms

One of the benefits of absorbing the COW features is that now we can freely use them as building blocks for additional functionality, inserting features 7 mechanics that would previously require some non-trivial Lua elbow grease. Realistic sub comms are one such example.

When this feature is enabled, submarines who go below 40m depth go off the communications grid. As in other cases of comms disruption, they are no longer under direct player control, and only their last reported location is available (this BTW means that sending them deep without first tasking them with a mission may be pointless, since they will simply sit there). The player can, at any time, send a “bell-ringer” ELF signal to a no-comms submarine to recall it to re-establish comms (right-click on sub icon, click on “Summon to re-establish comms”). The sub will attempt to rise to shallow depth to rejoin (this may not be immediate, as it will try to evade nearby hostiles). Once it rejoins, the sub will share its contact updates with its parent side.

Never stray from the path

One of the more persistent problem of CMANO’s navigation AI was that its pathfinder was binary-based: Either a spot was passable by a unit, or it was not. This resulted in a tendency in units to aggressively cut corners in their plotted paths in order to travel to their destination as efficiently as possible, sometimes coming dangerously close to shore or other obstacles. This can result in some unintended “beach surfing”, as this example demonstrates:

In this case, the Freedom is placed in the mouth of the Hormuz straits and ordered to travel inside the Persian Gulf. It plots an efficient path — too efficient, as it nearly clips the shoreline at Khasab. There is a possibility that the ship may get stuck on the shore, if left unattended. Ships generally don’t navigate this way if they can help it.

To address this issue, we implemented a new cost-based pathfinder that evaluates a location’s suitability based on a number of different factors. In the case of ships, a primary concern is local depth and proximity to terrain; generally large ships tend to prefer to maximize both, while smaller ships are somewhat more freewheeling. Using this new logic, ships are now able to plot much more realistic (and gameplay-friendly) courses, as the very same example played out in CMO demonstrates:

Stay in formation

This has been a popular request, which we are finally happy to oblige: The formation editor can now also been used to define and edit aircraft formations. It sounds like a simple change, but in reality we had to go through extensive reviews of the mission AI logics to make sure that this did not interfere with aircraft assembly & grouping logics (in some places it did, and we had to make adjustments). But we think it’s worth it, and you will probably agree: This opens up new opportunities for formation tactics and optimum placement & distribution of air assets in a group, both for regular transit or surveillance as well as combat.

A patrol of four MiG-31B sweeps over the sea north of Murmansk. The group is in loose line abreast formation, to widen the search area. Lead and the nearest wingman are already in position, while the two outer members race to catch up to their stations.

I like the way you move (and think)

Various tweaks & improvements have been applied to the kinematics models & AI routines, with a strong emphasis on aircraft and missiles. Among other items:

- Numerous tweaks to aircraft flight model, specifically for “combat” conditions. For example, aircraft no longer “wiggle” between headings as they must first roll towards the turn direction before committing to a turn. This in turn makes inherent roll-rate more important to close air combat maneuvers (This is easier to observe in the Tacview window).

- Helicopter diving rates have been reduced (they were previously dropping like a stone)

- Surface- and underwater-launched missiles now use the same improved pitch kinematics as air-launched missiles, resulting is more realistic trajectories.

- Lofted AAW missiles now begin their terminal dive earlier, to avoid a too-steep approach to the target.

- Air combat AI improvement: Aircraft now consider approaching fighters/interceptors as imminent threat, not just missiles. This helps AI-controlled aircraft perform more proactive evasive maneuvers against fighters about to perform gun attacks on them (e.g. MiG-17 vs F-105).

- Significant change in unit AI logic: The “evaluate targets” and “evaluate threats” logics are now not performed on every sim pulse, but instead on regular intervals dictated by the OODA-Targeting (modified by crew proficiency) and OODA-Evasion values respectively. Therefore these two OODA values, combined with crew proficiency, become even more critical to a unit’s effectiveness and survivability.

https://www.youtube.com/watch?v=tSh1Tk0-R5E&feature=youtu.be

Feeling the strain: Aircrew G-tolerance

If you have ever closely watched a dogfight in CMANO, one of the things that may seem odd if you are familiar with real-world ACM is the seemingly iron-man constitution of the virtual fighter pilots; they seem to be constantly engaging in turns, climbs and dives as tight as their airframes will allow them, without regard for their own fatigue from the continuous high-G acceleration loads. This isn’t how close air combat works in real life; even the best pilots need a breather between aggressive maneuvering in order to allow their bodies to recover, otherwise they become more dangerous to themselves and their side than to the enemy (yes UCAV fanboys, you can laugh now). So we set out to model this nuance in CMO.

When an aircraft is considered to be performing “combat maneuvers”, an extra UI element is shown on unit status panel:

This represents the crew’s tolerance to hard maneuvering. The longer the aircraft is continuously pulling a hard turn, the more this buffer fills up. (Getting out of a hard maneuver, e.g. reversing a turn, reduces this strain and allows recovering. This is one of the reasons that scissors are a popular practice in RL dogfights.)

Once the tolerance is exhausted, the crew begin suffering G-LOC and have to significantly relax the turn (this is easier to notice in Tacview, as the aircraft wing bank visibly changes, but you should also observe a noticeably larger turn radius on the map’s top-down view as well), while regaining their senses and control of the aircraft. Of course the aircraft is particularly more vulnerable during this recovery period, which makes cooperative tactics and mutual support all the more important. The maximum tolerance level depends on aircrew proficiency, so experienced pilots get one more significant advantage against greener adversaries (Ed: Like they need one more…).

Radar frequency agility

Not all radars are created equal. One particular feature that became quite important during the Cold War for radar performance against jamming and other countermeasures was its ability to “frequency hop”, ie. shift its operating frequency to a different value within its feasible operating range. CMO now models this, and two particular operational benefits of this ability:

- Frequency-agile radars are significantly harder to noise-jam, as they can shift to a new frequency once their existing one gets flooded with static. The jammer has to either start hunting for the new frequency, effectively hopping after the radar operator, or alternatively flood the entire frequency range with jamming power to make hopping pointless; both counter-actions are feasible but they both have complications of their own.

- These radars are significantly less susceptible to doppler-notching maneuvers (more on this below).

This feature became so important to radar engineering & operations from the mid-1960s that, quite often, the main difference between a very capable domestic radar set and its export-cleared variant was the physical removal of the frequency-hopping circuits in the latter kit.

As in other cases, electronic-scan arrays and especially AESAs get massive advantages; in this case, all P/AESAs are automatically considered as being frequency-agile, while older mechanical-scan sets have to explicitly have this declared in the simulation databases.

A notch on the bedpost

Doppler notching, a maneuver designed to exploit a fundamental flaw in pulse-doppler radars in look-down mode, has been included until now abstractly in CMANO as part of the weapon endgame calculations. This time, however, it can be actively and preemptively used as a maneuver, both for missile track lock-break and for general surveillance radar detection avoidance.

Aircraft can attempt to fly perpendicular to an emitter using doppler filtering in order to hide inside its “blind” velocity gate. The effectiveness of the maneuver varies with crew skill (an “ace” pilot will execute it far more effectively than a novice), to discourage manual micromanagement by players. Aircraft under missile attack with a doppler radar guiding the missile will also actively try to beam the radar instead of the missile (the geometry of the two axes can vary significantly). The maneuver is ineffective against pulse-only radars and less effective against frequency-agile radars.

Players can also deliberately plot courses for aircraft that fly perpendicular to known PD search radars, to reduce the actual detection range. (If you’re a Microprose F-19/F-117 virtual vet this may bring back some memories).

This option has two settings: “Random within area” (default) and “Repetable loop”. “Random within area” is the patrol behavior you are already familiar with. If “Repetable loop” is selected instead, the assigned units treat the reference points of the patrol area as a racetrack, just like in a support mission. This allows the player (or scenario author) to define precise, consistent patrol patterns (figure-8, expanding spiral, ladder etc.) while also retaining the “investigate & prosecute” behavior that distinguishes patrol missions from support ones.

When the “Repeatable loop” setting is active, the patrol area is depicted very similarly to support mission racetracks, to emphasize the difference:

Till death do us part: Merge scenarios

This is another gift from the PE world, and we bet it will be a much-appreciated one. If you are a scenario author, you are already familiar with the base & group import/export (ie. the .inst files), and how convenient and time-saving they can be when putting together the elements of a scenario. This “construct once, re-use everywhere” concept has now been taken a step further: You can make individual scenarios and then merge them into a single scenario.

Scenario #1 is treated as the “base” for the merge, so behind the scenes you are essentially cherry-picking all transferable items from scen #2 and adding them to scen #1 (the log output also makes this clear). Elements that are “atomic” or otherwise hard to merge (date & time, weather, title & description etc.) will, on the output, be those of scenario #1.

The power & versatility of this tool should be readily apparent to anyone who’s delved in scenario editing. Now you can create once, store and re-use not just groups or bases, but multiple foundational scenario elements: Sides, missions, events, actions, conditions etc. We have been repeatedly amused by the wacky, quite unforeseen ways the original import/export functionality has been employed by industrious scenario creators, so naturally we are curious to observe how they’ll bend this one to their will.

Lua loco

The Lua API continues its expansion in CMO and offers additional hooks into the simulation engine as well as various methods for pulling the strings of the running scenario. One of the new hooks ties directly into the AI model: You can individually instruct units to, quite literally, not think for themselves (You in the back, quipping “you mean they do this now?” – SIT DOWN!). More specifically, you can set individual units to skip their AI routines for evaluating valid targets and picking out the primary one among them. This has two direct benefits:

- It makes it easier to implement custom targeting AI routines in Lua, since an author not longer has to “compete” with the build-in AI for this.

- As these routines are among the most CPU-expensive pieces of the simulation pipeline, disabling them can have a drastic effect on the speed & scalability of a large scenario. For example, if you disable the AI cycles of all static/inactive buildings, then only “active” units will use the CPU for this work (internally Command already does a lot of such optimizations, but since it cannot “intrinsically” know which units are static & inactive, it has to check them, which itself is not free).

Additionally, this ability can allow simulating “dormant” states for units (e.g. units begin a scenario in a “comatose” state, but later because of XYZ they become activated).

Oct 22, 2019

DISCO ELYSIUM - Substances and You

Disco Elysium - The Final Cut - ZAUM_Dani

Now that Disco Elysium is out – and people seem to be diggin' it – let's talk about all the drugs you can do in it. Let's talk about Pyrholidon, The Lightning, Al Gul and the sumptuous Tabac herb. Think of this as a little designer-tutorial on how to get the best out of those bad boys. Or how to make due without them, because...

1. First, your detective doesn't have to do drugs. A straight edge run is a great way to play the game. (One favoured by our writers, for example). It makes the game more challenging and adds some extra role playing tension – in the form of temptation. After all, there is no temptation without abstinence.

2. Drugs kill. If you do go down that route, know this: drugs give flat bonuses to your main Stats. But they also damage your Health and Morale. Which can lead to heart attacks and giving up. This means you should at least aim to heal any damage immediately after taking a hit. Hands shaking from the Lightning? Smooth it over with some magnesium. Smoking got you groggy? Nosaphed will clean those sinuses!

3. Number one trick – drugs also raise the learning caps of your Skills. In Disco Elysium, your initial Stats decide how many points you can put in the Skills under them. Because drugs (temporarily) raise your main Stat, they also raise your learning caps. When the effect wears off, you get to keep the point you put into your Skill. (Fun fact: this started out as a bug, but we kept it because testers liked it.)

4. Most drugs have sub-types, or “brands”. (Cigarettes, for example, can be Astra or Tioumoutiri.) These are cosmetic. Just like in real life, brands do nothing. They're just random fetishism – speed is speed, nicotine is nicotine, wine doesn't kill you any less if it's expensive.

5. Drugs have charges. Electrochemistry gives you extra charges. The base amount of charges for all drugs is 3. If you have 4+ Electrochemistry at the time of first acquiring the drug, you get 4 charges instead; 7+ Electrochemistry gives you 5 charges.

6. You might wanna top up. One charge lasts 1 hour of in game time. Most big scenes take longer than that to complete. So if you blast your drug of choice before going in – in preparation, as you would a spell before combat – you'll run out half way through the ordeal. Amend this by further blasting in the thick of it! Bring that brewskie, that ciggy, or that anti-radiation drug to the lion's den by keeping it your held slot. Then keep an eye on the clock and take a preventive hit.

7. Containers with legal drugs (smokes and alcohol) are visible for all – while containers with illegal drugs (pyrholidon and speed) become visible only after first using the drug. There might be some surprises waiting in, say, Cindy's coal room. Or the fish market. But only for the initiated. If you're innocent you won't know what to look for.

8. There is one drug for each of the four main Stats in the game. You can use this to “fix” a weakness in your character build. Low Intellect detectives might find themselves smoking a lot, while low Psyche detectives have more to gain from doing Pyrholidon. Low Physiqe detectives are prone to sucking on a bottle. Even a low Motorics detective is stupid enough to try that jump when they're on the Lightning Rail.

Here is a summary on all the drugs in the game, where to find them and how they interact with your thoughts.

SMOKES

Bonus: +1 Intellect.
Damages: -1 Health (-1 Endurance)
Healed by: Nosaphed, Drouamine.
Get it from: Frittte, Rosemary, containers on the map
Brands: Astra, Tiomoutiri
Thoughts: “Boiadeiro” from passing Manana's Conceptualization check amplifies smokes.
Did you know: tracking the “Tioumoutiri” brand of smokes can lead to revelations in your main investigation.

PYRHOLIDON

Bonus: +1 Psyche
Damages: -1 Health (-1 Endurance)
Healed by: Nosaphed, Drouamine
Get it from: Roy, after passing an Electrochemistry white check.
Brands: none
Thoughts. “Cop of the Apocalypse” doubles Pyrholidon’s effect.
Did you know: Pyrholidon is an anti-radiation drug. Using anti-radiation drugs recreationally was a beloved pastime of Soviet hippies and punk-rockers. (Tareen was one such drug.)

ALCOHOL

Bonus: +1 Physique
Damages: -1 Morale (-1 Volition)
Healed by: Magnesium, Hypnogamma
Get it from: Fritte, Rosemary, from containers on the map
Brands: Commodore Red, Potent Pilsner, Pale-Aged Vodka
Thoughts: “Revacholian Nationhood” unlocks the full heroic power of alcohol. “Waste Land of Reality” does... the opposite.
Did you know: Disco Elysium is produced by reformed alkies who quit “the Ghoul”; turns out it's not possible to get a 60+ hour RPG and a drink on simultaneously.

SPEED

Bonus: +1 Motorics
Damages: -1 Morale (-1 Volition)
Healed by: Magnesium, Hypnogamma
Get it from: Klaasje's medicine cabinet, a quest after passing Cuno's Empathy check.
Brands: Preptide, trucker speed with a straw in it
Thoughts: “Lonesome Long Way Home” makes speed also give +1 Psyche too.
Did you know: The logos you see for drugs, including the one for amphetamine, are their actual molecules. (Or, in Pyrholidon's case, their fictitious molecules). We enrolled the help of a chemist to make sure they make sense. Thanks, chemist!

So that's it. Drugs are a multi-tool offering more flexibility and role playing options. But, on the other hand, they distort your personality and are completely incompatible with staying alive after you're 30 (trust us). And it's quite possible to make do without them. If there were a character sheet in Disco Elysium, the last field would be: Drug of Choice. The strongest DoC is the fifth, hidden drug we like to call Life and Your Mother's Love.

True bad asses get high on that :)

Until next time,
Roberto Kurvitzo, maker of RPG's; reformed Narcomaniac.

PS. Thank you for playing and sharing your stories in forums and media. It's you who make it all worthwhile.