Alexander Wales

megaword author, book scarfer, amateur research fiend

[alexanderwales]

The Index

This is an index of things I've written and posted online, with minimal descriptions because most of them have blurbs if you click the link. This list is not exhaustive, especially because there are a bunch of short stories and dribbles in various places. If something you liked is missing, let me know.

Web Serials

Worth the Candle - Juniper Smith is a teenaged Dungeon Master who ends up in a world filled with all the things he dreamt up for his campaigns, along with signs of his friend who died months earlier. This Used to be About Dungeons - Five teenagers live in a house together, bake bread, tend the garden, and occasionally fight monsters in dungeons. Thresholder - Thresholders travel from world to world, fantasy one minute and scifi the next, always encountering an opponent, growing stronger as they battle. Shadows of the Limelight - Fame gives you superpowers, and Dominic just saved the world's greatest hero from defeat in full view of a large audience. Glimwarden (unfinished) - A small town huddles around lanterns that keep the darklings at bay. Four teenagers must grow in power as the darkness encroaches. The Dark Wizard of Donkerk (unedited) - Two men steal a baby from an orphanage, then find out he's too cute to sacrifice and raise him as their own.

Fanfic

The Metropolitan Man (Superman) - Lex Luthor attempts to unravel the secrets of the alien. A Common Sense Guide to Doing the Most Good (Superman) - Superman gets really into effective altruism. Instruments of Destruction (Star Wars) - A fable of project management aboard the second Death Star, through the eyes of Admiral Tian Jerjerrod. Branches on the Tree of Time (Terminator) - Sarah Connor is working as a software engineer at UCLA when a naked man shows up on her doorstep. A Bluer Shade of White (Frozen) - Elsa can make life, and Olaf is smarter than he looks.

Shorts

Eager Readers in Your Area - Artificial intelligence has left authors scrambling for readers. Charlotte clicks on an ad. Variations - An orc visits an art exhibition where she feels out of place. Contratto - Julia takes a job as a marketer, working for the vampires to keep their secrets safe. The Randi Prize - James Randi offers a prize for anyone who can demonstrate supernatural abilities. Coming Home - After a long time isekaied to a fantasy kingdom, an errant father has coffee with his estranged son. Drips and Drabs - A collection of microfiction

I also post short stuff to this very tumblr, which can usually be found under the #microfiction tag unless I forget. Usually this is mirrored on AO3, unless I'm lazy.

Web Comics

Millennial Scarlet - Lamont Pearce is a gig economy demon hunter whose mother ran a government agency meant to defend against Hell. Worth the Candle - A webcomic adaptation of the web serial

Non-Fiction

The AI Art Apocalypse - Slightly outdated thoughts from 2022. Why to Write a Sex Scene - Observations on the narrative purpose of carnal pursuits. Game Review: Underhill - This review contains no screenshots, because this game does not exist. Writing: An FAQ - Accumulated wisdom from 4 million words and counting. Creating Interesting Magic - A much-requested post on making interesting magic systems (and characters, and plots, and worlds). How to Write a Web Serial - It's both easier and harder than you think. The Trouble with Writing Nazis - On giving villains too much credit. Interesting Things to do with Time Loops - Exploring the boundaries of the conceit.

[alexanderwales]

Helping a friend move, and I am actually getting too old for this. When I was 20 I could move a couch like you wouldn't believe.

[alexanderwales]

"You got that couch in there all alone? How?"

"I dunno, I just ... moved it? It wasn't that heavy."

"But I measured it! It wasn't going to fit! Did you take off the feet?"

"No, I just turned at an angle to get through the door."

"Wow. Have you ever thought of doing this professionally?"

"No?"

"Regionals are coming up, I think you could do great. But the field is pretty competitive. My dad was a professional mover before a career-ending injury, he taught me all about it."

"Well, I guess if I have a gift for it ..."

This Fall, Tune in For the Hit New Special Interest Anime: Moving Up!!!

[alexanderwales]

Ugly babies in the robin's nest

[alexanderwales]

Helping a friend move, and I am actually getting too old for this. When I was 20 I could move a couch like you wouldn't believe.

[alexanderwales]

I sort of want to play a game where you're a hiring manager, trying to weed through applicants to find the perfect person for a position, though I don't entirely know how you'd gamify it. In a lot of management games, you see all the perks, buffs, and debuffs your employees have right away, but maybe there's a fun version of this where you find out their quirks and flaws as they work for you, and sometimes you have make a choice whether to let someone go or silo them off from others. Maybe this is my dream SCP game, where you're corralling unethical doctors, psychopathic guards, people who are inducted because they were in the wrong place at the wrong time, etc., attempting to make the best of a bad situation when it comes to hiring people onto the project.

[alexanderwales]

I was reading through the U.S. Copyright Office's "What Does Copyright Protect?" page, mostly to make sure that I wasn't being wrong on the internet, but I absolutely love that this is part of their FAQ:

How do I protect my sighting of Elvis? Copyright law does not protect sightings. [,,,]

There are only 12 questions on the FAQ, and this is one of them.

[alexanderwales]

One of the mechanically interesting things about 007 First Light is that they have a "License to Kill" built into the game, and how they decided this works is that you can't just shoot people in the head, even if they're armed guards.

However, if an enemy shows "lethal intent", by e.g. drawing their weapon, then you do have a license to kill. So there are situations where you just escalate the confrontation until someone is pointing a gun at you, and then you can start sniping people in the head with abandon.

I sort of understand the gameplay reasons for this, but it's hilariously awful when viewed through the lens of any kind of realism or theory of state violence.

[alexanderwales]

My son is reading through Animorphs, and while I got him the first six books as a box set, apparently the later books are just not that easy to find. We got the seventh from the library, but they don't have the complete collection, and I guess after having been at the Scholastic Book Fairs or whatever, they just never got reprinted? They're available for truly exorbitant prices on eBay, not available on Amazon except as ebooks, and I guess we're just going to have to set him up with an e-reader and pirate them.

Makes me sad to see a staple of my childhood vanish like that, so I might have to put out a call to see if any of my friends held on to their collection.

[alexanderwales]

Given that llm summary and retrieval are an ever-more-popular way of finding things on the internet, might it be worthwhile for people doing web-publishing to have simulations-or-instances of the most-used models (Google search ai, free ChatGPT, free Gemini, free Claude) produce summaries of their works, so the producers can learn how the machines take their work?

If I understand you correctly, you're asking whether authors should have the LLMs ingest their works and look at the output to make sure that the LLM is correctly summarizing it?

I guess that's sensible, if that works.

But also, the LLMs are not actually reading the work, they're reading the first page of search results or possibly the first chapter (if that shows up in search results). So instead, an author that's focused on getting their work out there should be thinking about the blurb, or how third parties will describe the story, because this is what the LLM is going to synthesize.

And you should already be calculating the blurb to immediately signal what kind of story it is, if you're audience conscious. It should be calibrated to the lowest common denominator, whoever you decide that is for your story. And I don't think that's going to be substantially different from what you'd do regularly.

You absolutely should not be writing for the LLM, because the LLM is not going to read your story unless explicitly told to.

[alexanderwales]

When Black & White came out, they had creature models that changed based on whether it had muscle, was eating a lot, young or old, etc., blending the mesh or something to achieve the effect. Then in addition to that, they had the animations be partly procedural, so that they would mostly look correct even if the physical size of the creature was different.

I always thought that this was the future of videogames and animation, but we're only just barely getting to the point of animations that are driven by limb length and underlying physio anatomy. In theory, a character should walk different depending on their musculature, what kind of load they're carrying, whether they have an injury, etc., and this should all be dynamic.

But the modern AAA games have gone in the direction of blended mocap and a singular character rendered in very high fidelity that I guess wouldn't make sense to have dynamic appearance and animation for, and the indies tend to use whatever is available out of the box. Plus if you're going to allow characters to be different heights, weights, etc., you start growing concerned about whether there's some optimal thing to be, and certainly don't want players to optimize into picking the shortest dude they can so the hitbox is tiny.

So the technology and systems that seemed to hold a lot of promise 25 years ago have fallen into a sort of developmental no man's land.

[alexanderwales]

I started on 007 First Light's "TacSim" stuff, which are "virtual reality training levels", AKA an excuse to reuse a bunch of assets with more gameplay and challenges. They're fun, because the core gameplay is fun, but they really do make me think about how this conversation would have went down at MI6.

"Bond, it seems here that you used the Tactical Simulator to see whether you could throw ten men off the edge of a platform without going more than thirty seconds between each of them?"

"Yes, ma'am."

"And you spent almost an hour on this."

"Yes, ma'am. It was for training, ma'am."

"Do you anticipate that you'll need to throw men from ledges often?"

"Not just throw them, ma'am, but also activate convenient fans to blow them off. Yes, ma'am. And to practice the one-liners. 'Didn't seem he was a fan of that' or 'that really blew him away'. It's important to get the reps in."

"Ah, well, if it's for the one-liners, carry on."

[alexanderwales]

I finished 007 First Light, the new James Bond game, and while I liked it overall, it was definitely a James Bond game that made me feel like Nathan Drake or Agent 47, depending on the scene, only rarely James Bond. I feel like it's got a lot to love, though I do wish that it hadn't been essentially a prequel/legacy/reboot game. You spend the whole game in the shadow of the previous 00 program, and only really get established as a proper respected spy in the final moments, so you're less "James Bond, superspy" and more "James Bond, recruit who has to prove himself".

[alexanderwales]

The anthropologist's barely disguised fetish.

[alexanderwales]

The only Factorio achievement I don't have is the one for finishing the Space Age expansion in under 40 hours, and occasionally I feel the urge to go back and complete it, which surely can't take that much time, since 40 hours is a rather comfortable time limit. But one of the ways to complete Space Age fast is to have a bunch of blueprints ready to slap down instantly, and those would have to be done on a separate save file, which I think is where the amount of time needed would rapidly increase.

Though I suppose enough time has passed for there to be a readymade blueprint book of builds to place ...

[alexanderwales]

I spent some time today working on a little toy model game, wherein you have to arrange "ship parts" that are made of squares, paying attention to weight distribution, center of mass, aerodynamics, etc. It was a bunch of vibecoding to test whether the shipbuilding was actually a fun thing to do, something that wouldn't have been possible to do rapidly before the LLM, at least not with the current rust on my programming skills.

It's pretty fun, but would require a lot of work to add any amount of balance, especially when The Vision is that it would be a roguelite part-drafting FTL type of thing.

A few things that occurred to be during the prototyping:

It's fun to have two different modes for construction, "bolt" and "rebuild". When you're out and about, flying from planet to planet, you're using "bolt", attaching parts to your ship as well as you can, jettisoning things you no longer need to make space, trying not to unbalance it. But then you get to some kind of rest point, and you can take the whole ship apart to engineer it into the best version of itself. Like healing in Slay the Spire, but you're getting in the guts of the machine!

There's something really pleasing about polyominos, which is that they increase in complexity as you add in more squares. There are two shapes you can make with three squares, seven with four squares, and twelve with five squares.

It would be interesting to have the polyominoes have mechanical identity based on their shape. Storage is O-shaped, power is S-shaped, weapons systems are L-shaped, thrusters are T-shaped ... the problem with this approach is that if part of the gameplay is ship-balancing, we don't want there to be a "standard" ship. So probably three of the shapes are "core" predictable ones, cargo can be any shape, and then the other non-core functions? Mostly I like the idea of going from tertomino T-shape to pentomino T-shape.

Tiny little one-square or two-square shapes are good here as a method of balancing out the ship and making sure that aerodynamics are good? I'm thinking of them like little bonus pieces you're almost always happy to gobble up.

Shipbuilding will depend on tension, and if there's "aerodynamics", mass balancing, center-of-thrust vs. center-of-mass, then there can naturally be archetypes that arise from that, like aerodynamic "missions" or "events" or whatever, which you are either optimized for or not. All this would be for TBD gameplay, but I was working on the bones.

Center-of-mass, center-of-thrust, and aerodynamics are all somewhat unintuitive, particularly if the pieces have different weights to them, so this needs to be very clear in the UI when you're fiddling with ship design, super reactive to the player moving pieces around.

Aerodynamics is handled via rise-over-run along the left and right edges, and I don't know how to calculate that and show that calculation in a way that's instantly legible.

Once the bones are done, there are a lot of neat events you could do, things like adjacency buffs, debuffs that inflate a piece by a square, miniaturization that makes a piece smaller (but might unbalance the ship until reconfiguration), and stuff like that.

These are mostly notes for my own benefit, as I don't intend to throw away more time on this. Vibe coding and prototyping is nice to see the shape of a thing, but the actual work of game-making is a lot of testing, tuning, etc. To actually have a collection of rules for how ship parts are generated, and a long list of events, takes much more than minimum-viable design.

[alexanderwales]

Four thousand years from now, people will debate whether and how the first software was even built, as surely we couldn't have done it on their own with no AI assistance. Must have been aliens.

[alexanderwales]

I'm a fan of puzzle games, and there are a few rules that are very common in the Standard Puzzle Game, like "you don't need to guess" and "there is only one valid solution".

But I was playing one of these puzzle games recently, and it had an additional stipulation, "you cannot use the fact that there is only one valid solution to complete the puzzle". And it was very funny to me, because yes, this makes sense to include, you don't want there to be a situation where a puzzle's solution depends on the assumption that there's only a single valid solution.

Seems like a headache to design though.