"Though the hardbrain segments you will use as the locus for most hexes have little in common with the brains of mages and beasts, there are certain specialised segments which do present a remarkable similarity in function, though not in form." This was spoken in the crisp tones of Anett He, my last lecturer for the day. Anett wore a robe of a purple so dark it looked black until its shiny surface caught the light. He's eyes were hidden behind massive dark, round lenses and he's face was obscured by a veritable map of scars.
I had not read ahead for Anett's class, so I had good reason to pay close attention.
"While we tend to refer to all hardbrain segments as 'demon brains', this type are more brainlike than most. They seem tailored specifically for storing facts as an interconnected web of weights and biases, rather than directly represented by exact sequences of fliplines. These 'brainwebs' are highly versatile but are even more opaque to our understanding than the workings of hexes. Therefore, we do not directly control brainwebs through hexing. Rather, we must train them similar to how a carer trains a young child."
This was all very useful information about a deeply hidden side of magic. It might not help with my project — attempting to take this into account would vastly complicate my work — but it was important to my understanding of demons. If I understood them, I might not fear them so much.
"Through secondary hexes, we can encourage these brainwebs to learn. The results are quite variable and depend significantly on the quality and structure of the learning material we provide to them, as well as the precise setup of the learning process. Understanding this well enough to produce good results and know the results are good will be the work of many years."
That was definitely something my project could not fix, though it might make the process a little easier. Streamlining the process of building hexes would save a lot of time for a lot of people. If I could make my hex-building hex work as well as I hoped, it would be easier to understand, prove and explain what a hex written in my human-readable structure actually did. I had to make it actually work first. Once I got some parts of it functional, I should be able to use the hex to create the hex.
"Here is a mathematical description of the learning process..."
I could look that up when I needed it, which meant I could get away with checking the logic of the latest parts of my design for a few minutes. The first step had been to decide upon the logical building blocks I wanted to use to express the operation of a hex. Then I had to use those building blocks to write out an exact description of how to turn any arrangement of those building blocks into the corresponding working hex. Then I had to check that it was actually correct, through much close examination. Then I had to meticulously turn each part of that description into a hex by manually following those steps myself.
Once I completed a self-contained unit of the hex, I would be able to enact it with my scryer to see if it practically did what I expected it to. Then I might find that it did not, and would have to determine why. Perhaps the problem would be with my design. Perhaps the problem would be that I built the hex incorrectly. I did not expect it to be at all obvious which or where, and strongly hoped to avoid that task.
I was still working on step two and occasionally revisited step one. It had taken me a lot of heavy reading and pondering to understand what convolutions I would need to put a scryer's hardbrain through to achieve my goal. Without my practice in envisaging the logical flow of the Codex and my eerily good memory, I wouldn't have had much hope of getting even this far. But I wasn't far enough yet, and even my concentration techniques could not completely prevent my brain from tying itself in knots.
I tapped on my scryer to activate the hex for reading and writing sections of text, which I thought of as the magical equivalent of a notebook. The viewing pane adjusted itself to show a list of my notes, which were organised in an imaginary 'folder'. Or rather, it was a magical folder. It wasn't physical, but it wasn't unreal either. There was no actual folder but it operated according to the metaphor of a folder and was marked with a folder pictogram. Magic could be confusing.
Metaphor or otherwise, I could see the names I had assigned to my various notes in my notes folder. For complicated and uninteresting reasons, the notebook hex was not well suited for reading or creating hexes. There was a separate hex for that purpose, called 'hexedit'. The students had been repeatedly warned not to use hexedit to edit any hexes that came with our scryers, as that could cause serious damage and make a scryer unusable. We were also warned against making hexes that modified other hexes, as that could cause the same damage, possibly on a wider scale. That was something I had to watch out for, though it was unlikely that I would do it by accident during the process of writing a hex for creating new hexes.
Some of my notes were design notes for assignments and my own projects, some were lecture notes and still others were my observations on mage life. I had also taken extensive notes on my adventure with Skids and the events surrounding it. That was too important to forget, and there was much I did not yet fully understand. I also had some notes on my attempts at understanding the data Skids had duplicated from Sente's relic, the scryer-like object believed to be from prehistoric times.
I had not had any success at doing anything useful with the duplicated data. It had ended up as one massive chunk that my scryer treated as a single piece. That piece was so big that the only way I was able to look at any of it was to duplicate a small chunk with a simple hex of my own design, and then use the notebook hex or hexedit to read it. That was like exploring a mine larger than the entire hive by chipping off a little fragment of rock with a hammer and putting it under a microscope. It was close to impossible to find anything specific, and about as difficult to know what I was looking at.
Anett was still talking about hexes for training brainwebs. As far as I could see, the rest of the class was paying close attention.
None of this would help me in my project. I didn't want to teach demons to write hexes. Especially not using a process that produced highly variable results. With my method, I would be able to see and check the logical rules the hardbrain followed to produce hexes. Teaching a brainweb would be like trying to teach a student who didn't speak, by showing examples and hoping he got the right idea. And proving that such a hex actually did the right thing would be a nightmare.
I quickly rejected that entire concept and returned to my notes. On a whim, I opened the relic data note. I had recorded the locations in the data where I had found some readable text. Well, readable was somewhat aspirational, as I could not actually understand it. I had also found and located a few images. It had taken a lot of trial and error to find the start and end of each image in order to extract and view it. Unfortunately, the images had been nothing more than mystifying caricatures of facial expressions. They were amusing, but not significant or helpful.
There had to be something I could do to speed up the process of extracting useful pieces from the big blob of data. I felt there was a way, but I had been so busy with assignments and my hex generation project that I had not given it much attention.
That was when an idea hit me like a tunnel collapsing, and I could see nothing but pieces of the idea surrounding me. A significant part of my big hex was responsible for traversing through a stream of data and picking out certain words and patterns. That was the one part I had actually made work, but it was not of much practical use on its own. Except...
I had kept each part of the big hex organised separately, with extensive notes on how every part functioned and how it corresponded to my readable design. That gave me a chance of actually progressing instead of giving up in the face of an almost unreadable mess of symbols. Now, it let me repurpose that region of the hex in only a few minutes, instead of taking hours and needing to renumber everything that referred to a specific distance from the start of the hex.
As Anett began giving examples of how training a brainweb might go wrong, I completed my pattern-searching hex. In order to use it, I had to decide what I was looking for. Should I look for words? Looking for specific familiar words was probably a waste of time, but looking for any words might get me somewhere. There would probably be a lot of words in such a huge chunk of data, more than the notebook hex could handle.
Hmm... the hex...
I wiggled my fingers in specific patterns to choose the symbols I wanted. It had taken me months of practice to learn how to do that, but it was faster than operating the levers on the scryer. It was slower than a mage with two flesh hands could manage, but I could only work with what I had. The symbols appeared on the scryer's viewing pane: 7f454c46. I wasn't searching for text. I was searching for hexes.
"Chloe! Are you with us? Did you drift away from my lecture again?"
I whipped my head up from my scryer. "Just going over my notes."
"Then tell me this: what is the biggest factor to successful brainweb training?"
"The quality of the learning material?" I'd been listening to that part.
"Are you certain?" Anett asked, raising a finely sculpted eyebrow. The eyebrow was augmented with a tattooed explosion of curly violet lines.
"Quality and... structure! Yes, structure."
"And what does that mean practically?"
"Um..." I scrambled to think of a sensible answer. "A brainweb will learn the simplest patterns it finds, which might not be the patterns you want it to learn. If you were teaching it to identify pictures of tools held by mages, instead of actually considering the shapes of the tools it might... er... look only at the colours of mages' sleeves."
"That's a... creative explanation. But not an incorrect one."
I barely restrained a loud sigh of relief.
"Our time today is up, but please no one hesitate to see me after if you have questions about your assignment. Class dismissed."
I looked back down at my scryer, ready to pack it away. It had found hundreds of hexes in the relic data already. Or rather it had found that many magic numbers identifying the start of hexes. It remained to be seen whether they were actually hexes. Prehistoric hexes.
"Hey Chloe. That was pretty interesting, wasn't it?" It was Spike speaking, from one desk to my right. I hadn't seen him.
"Yeah. Pretty interesting," I parroted as I fastened my bag.
"So what were you really doing?"
My heart leaped, like our train every time Skids disengaged the aetherspinner. I'd been caught. "Huh?"
"You were hardly listening to most of that. Were you starting one of your assignments?"
My fear turned to anger at he's condescension. "I am not! behind! on my assignments!" Most of the students were leaving as quickly as possible, but a few strugglers turned to look at the commotion. I was past caring.
Spike stood and stepped closer in order to speak more quietly. "Hey, easy there, it was a logical guess to make."
"Logical? You think me being behind is logical based on... what exactly?"
"Well you are a... an older student."
My tone turned mocking. "Really, I'm 'older'*'. Anett is 'older'."
"Anett isn't a student. I meant... You're a late start. And you were having trouble remembering your assignments earlier tonight."
"How do you know all my assignments, Spike? We're not in all the same classes!"
Spike smirked. "It's easy to look up the assignments for each class with a scryer."
"But why?"
"I thought you might need..." He stopped himself but it was too late.
"You thought I might need help? Because you think I'm not a smart as you, don't you? Well I've already finished all the assignments, so maybe you are the one who needs help!"
"Um... If you're offering?" Spike asked, sounding hopeful that I wasn't really upset.
"Well too bad, I've got far more important things to do. Like making a new hex that can make hex design actually sensible so we can stop wasting time with the repetitive and mind numbing process of turning understandable logical concepts into unreadable groups of arbitrary symbols you all need to memorise!"
"You're being very loud, Chloe," said Anett. "I may tolerate some unrulyness from my younger students, but you are not one of them. And you're wasting time in my class with a pointless project that's doomed to fail. If you have so much extra time and wish to be welcome here after that outburst, then I can speak with Gennn about setting you some extra practice exercises. How does that sound for a logical concept, hmm?"
I wisely but belatedly kept my mouth shut.