Since repairing research capabilities to continue my design was going to take a while, it was time to decide on manufacturing and maintenance priorities.

Suppressed // Mindfulness: Extreme subversion of will detected (possibilities: 99.4% external subversion, 0.5% intentional action, 0.1% benign design). // Suppressed

I had enough chicken mechbots. Besides which they had been less useful than I had originally hoped, I actually had plans to disassemble a few for parts for my new design.

Security in general was no longer a major priority. The main security tunnels between the known surface entrance and the annex were repaired—which was implied to be only 10% of its capability when operated autonomously. Perhaps if I had Manufactured Workers the existing tunnel designs would work well. While I did still need to install gearways in the tunnels, there wasn't much more I could do at the moment.

I did have obligations now, and if I dedicated all of my manufacturing time towards battlemechs I could fulfill them, just barely in time. The better strategy would be to increase my manufacturing fivefold and do it all in a week.

Maintenance was going to be the next priority then. This would also allow me to repair my research and manufacturing facilities, as well as improve security with gearways.

That also meant I was going to need more power, I was already pushing the limits of my current generation. I could ignite more reactors, but the ones I had still needed to be tuned. A process which would take a significant amount of my maintenance bots and computational power.

What I really needed were more maintenance mechbots. They had been my first priority upon awakening for a reason. I had already begun the manufacture of spare parts for the maintenance mechbots the moment the battle was concluded, as there were still plenty I could repair. But now I began manufacturing some new maintenance mechbots. That, along with synergies—including allocating a new one for design performance—was my manufacturing plan.

Before my final decision on a maintenance plan I should probably rank up maintenance as I had planned.

Maintenance 1 • Maintenance Rank Upgrade Gain +2% performance on all maintenance tasks. Choose one of the following approaches to maintenance to empower. Reactive: Empowers repairs, providing a +25% task performance bonus to all repairs. Preventative: Empowers regular preventative maintenance allowing it to increase the value of an attribute to 110% of its maximum. (This uses no extra material.) Predictive: Empowers the use of monitoring to improve the prediction of maintenance events, preventing all repair and improvement maintenance downtime.

Preventative was very tempting, it was effectively Inner Perfection—which it should stack with—but as a rank upgrade rather than a perk. But I needed repairs now. Rank Upgrades weren't covered by Self Modifying Soul (which covered Perks) but I hoped I could change this decision later. I chose Reactive.

Maintenance 2 • Maintenance Rank Upgrade Gain +2% performance on all maintenance tasks. Choose a Maintenance Specialty. When performing tasks related to a Maintenance Specialty double the Maintenance rank bonus.

Following the outline of the manufacturing rank upgrades, this was as expected. I noted however the difference between a focus and a specialty in the wording. The choices were different too, here I could choose an Attribute—Research, Security, Manufacturing, or so on—or a physical substrate grouping—Municipality, Facility, Vehicle, or Object—notably I could pick everything covered by Craft as one group or the things covered by Build individually.

A new group of maintenance specific Mastery Perks had been unlocked with this rank, but they applied to physical substrate specialties, and weren't useful enough to change my preference.

I chose Manufacturing, obviously. Not just for the obvious reasons, it was also two thirds of everything that needed to be repaired, and I had a suspicion about the next rank upgrade.

Manufacturing 3 • Manufacturing Rank Upgrade Gain +2% performance on all maintenance tasks. Reactive II: Instead provides a +50% task performance bonus to all repairs. This bonus is doubled for applicable Maintenance Specialties.

Again, as expected. At Rank 3, the Rank 1 choice got better and involved the Rank 2 feature. I was pretty confident that Manufacturing 4 was going to be another Specialty.

One that I didn't really need at the moment. I decided to stick to my original maintenance plan: maintenance, research, and installing gearways in parts of security. Now with 50% more performance. But I decided it was better to rank up something else with my last free point.

The question was what to rank up. Research was tempting, but going by amount, Storage, Computation, Security, and Power were all more useful as choices.

In the end I chose Computation as a self improvement.

Computation 1 • Computation Rank Upgrade Gain +1% effective memory and +1% effective operations. Choose one of the following approaches to computation to empower. Processing: Empowers fundamental predictive processing architectural elements designed to decrease total time for operations to complete (this includes memory caches, branch predictors, and others), improving their accuracy to at least 90%. Storage: Empowers storage controllers at all levels providing them with embedded compression and encryption schemes that will not affect normal performance. Communication: Empowers routing and transmission capabilities connected to this computation infrastructure, all communication within the physical substrate will always replace either routing delay or transmission delay with the minimum possible value (evaluated per segment, the choice of which is replaced is always based on maximum benefit).

Certainly the most understated rank upgrade yet. Also it seemed oddly specific compared to the others.

Having now confirmed the pattern, these should improve again at Rank 3, and the improvement should interact with the system of focus and specialty that Computation had. Which was important to keep in mind, as none of these were particularly useful at the moment.

I could see the promise in Processing—the real bottleneck to my performance was not raw operations a second, it was how long I had to wait for data from my memory to get there, which my fundamental architecture mitigated by having predictive elements to guess what memory it needed to keep around—but my cache hit rates were already at 96%, and my branch predictor's accuracy was at 94%. Storage was obviously useful, just not to me, I had massive parallel arrays of storage with excellent encryption and compression schemes that put very little strain on my ability to store and retrieve information. Communication was the only instance of something that would actually help right now, even if it was a marginal benefit on the order of a millisecond.

I threw storage right out. Whatever the upgrade was, whatever it's interaction was, improving my ability to store information was in no way an important factor to me.

Between Processing, which I outclassed already, and Communication, which was pushing the limits of reality, there was a notable dichotomy between 'thinking the answer' and 'using the answer', like a human's brain and their nervous system. When it came to thinking up answers I already had an extra card up my metaphorical sleeve (would that make the soul cheating? human idioms were weird) with Thought Burst. And also considering an immediate benefit, however marginal, made it obvious. I chose Communication.

"When are you going to torture me?" asked Rubye Babington, the more responsive of my two prisoners. It was not an answer to the question I asked. She had become more agitated as my questioning had continued.

"I wasn't planning to. Would you like me to?" I replied.

"No! It's just… you have been asking very specific questions about how blood violently leaving my body would make me feel. For what must be hours. And it fucking seems like you are getting some sort of fucked up torture foreplay in," she was practically yelling now.

"I had been planning some practical experiments. I believe humans can survive significant blood loss."

"Fuck that. I'll tell you whatever. Look, it's not like I am a revolutionary ok? I was living in a shithole, even for the Racket, doing shitty things to be able to eat. You know The Republic hasn't taken refugees for years, and I obviously don't have money. So yea I signed up with Walla, pretty sure he gave me acolyte powers just to bed me, you want me to sign up with you? I'll do whatever just give me food and a place to sleep, and don't take my fucking blood", she was clenching her fists now, definitely a draft too.

"I see now how that could be considered torture," I was confident my models were improving, even if I was doing a lot of adjustment for source biases and getting conflicting answers from the two prisoners, "Just a few more questions then."

She nodded, apparently relieved.

"Do you feel that the removal of another human's blood in your vicinity would be torture for you?"

"The fuck is wrong with you man", that's a probable affirmative, "I thought you meant like camp details or whatever," she was back to yelling, "why are you even asking instead of just doing your fucking experiments?"

"It would be immoral to torture you."

Rubye started screaming. I wasn't quite sure how that response fit into my models.