The experiment with Jed didn't end up working. He did not become an agent. Though he did end up claiming land on my behalf—something I determined by carefully tracking the incoming claims and his movements through my mechbots. Which implied to me that he would have to take an active role in defending or attacking claims to count as an agent.

You have claimed-in-fact enough territory to qualify for one (1) additional soul perk.

You have claimed-in-fact enough territory to qualify for one (1) additional soul perk.

I had claimed nearly 600,000m² of prairies around the remnants of the barn. I stopped there for two reasons.

First, my chicken mechbots were now walking half a kilometer away which was putting them, with the addition of the security tunnels, near the maximum round trip distance. They were also having to walk in zig zag patterns to claim the land without gaps, something that was especially annoying as their claim radius seemed to shift for reasons I couldn't solve. My soul didn't exactly give me a map, so I was spending an egregious amount of computational power, moving hundreds mechbots in start and stops, and tracking Jed, to create one from a single truncated value.

Part of me just wanted to be done with the exercise—this much compute time wasn't something a factory spent on a task like this. I did my best to regulate those meta-heuristics. I needed to be more than just a factory. Besides, I had the compute time to spare.

The second was that I wasn't quite sure how this would be perceived by others. I had to assume that Walla would notice the change in his soul, or perhaps even get notifications like the ones I was suppressing. If he did notice, he would have to view this as an attack. I was worried he would call on fellow gangs if I pushed too hard, where as he might just try to solve it himself if it seemed minor.

There was also a certain amount of math that still wasn't adding up. My calculations for claims to tier had been correct (logarithmic base 10). But according to the map Obadiah had provided of Walla's territory, he should be Soul Tier 7, normally called a Raider when uncivilized, not called a Gang Leader as we was, which my counting put at Soul Tier 5. Perhaps this was linguistic drift, or perhaps my counting of uncivilized Lord names was wrong. While I was glad it wasn't the tier formula I was wrong about, the source of the inconsistency was unknown and that caused me to temper my advance with caution.

I also wasn't sure what was going to happen when Walla took back the territory.

This left me at Soul Tier 4. The first thing I did with the claims as soon as I got them was allocate as much as I could to my upgrade point generation. Next was picking more perks.

The Factory Intelligence

Tier 4 Primary Substrate: Facility (Level 12) Claims: Caverns: 0 / 1,200m² Prairies: 100,000 / 586,900m² Agents: Obadiah Cogsmith (Level 10), Elvira Rexword (Level 11), Mirabell Leeford (Level 10), Lawrence Cogsmith (Level 5) Perks (2 pending): Self Modifying Intelligence, Primary Substrate Mastery, Intelligence Core Primary Substrate Mastery (Tier 4, Facility) Upgrade Points: 0.2 (+0.8/day) Perks (2 pending): Specialization [Manufacturing], Focused Manufacturing Synergies Intelligence Core Perks (2 pending): Powerset Analysis, Thought Burst

Beginning with the perks I already had.

From mastery I really just wanted more immediate benefits. Which meant more upgrade points or, failing that, the ability to accelerate repairs (if one ignored storage I was barely level 8 in my current state). I saw only one poor avenue for directly improving Maintenance.

Powered [Attribute] • Mastery Perk Requires: Attribute Rank 4 Gain access to all upgrades for the Attribute which can be powered by your current claims.

I had seen the Powered Upgrades for Manufacturing from the elements Caverns gave—and now Prairies—after taking Specialization. Assuming allocation rank used the same logarithmic scaling that upgrade points did, I had not been impressed.

Earth Shaping • Powered Facility Upgrade Requires: Earth Claim, Manufacturing, 5 upgrade points per max allocation rank Using elemental earth, improve the manufacturing throughput of all processes that shape earth (clay, dirt, sand, gravel and similar substances) by +8% per allocation rank. Fire Furnaces • Powered Facility Upgrade Requires: Fire Claim, Manufacturing, 5 upgrade points per max allocation tier Using elemental fire, improve the performance of furnaces by +12% per allocation rank.

From this pattern it appeared they would be too specific to be helpful for Maintenance. Even if I had the upgrade points. I would have to experiment when I had some to spare.

I considered Rank Mastery and Inner Perfection instead. I immediately dismissed Inner Perfection as scaling my current values might help, but then I would have even more to repair. Better to get it later. Rank Mastery was more appealing—though I would have to avoid using it's free ranks for Manufacturing, out of an abundance of caution, due to potential interactions with Specialization—as I could use it on my Maintenance attribute.

I chose Rank Mastery and kept a point free incase Ranking up Maintenance gave me access to a better perk like Manufacturing had.

When I was first picking Intelligence Core upgrades I had been under a time crunch and attempting to pick something immediately useful. However the majority of the Intelligence Core perk catalog provided more interesting avenues for improvement, often focused on me as an informational entity.

Mindfulness • Intelligence Core Perk If not already a feature of your intelligence (you have 97% of this feature), gain the ability to inspect your own thought processes and gain an awareness of their shaping. Provides a soul-bound method, regardless of your current process, for this inspection. Provides a reflexive protection against mild external influences to your thought process, and a notification of most others. (Includes all directly connected external computational hardware.) Mind Merge • Intelligence Core Perk Requires: Copyable Intelligence Substrate If not already a feature of your intelligence (you have 2% of this feature), gain the ability to merge your intelligence with copies of it. Provides a soul-bound method, regardless of your current process, for merging. Provides predictive protection against the merge causing mild drift and subversion, and a statistical breakdown of risks otherwise. Modeling Accuracy Report • Intelligence Core Perk Requires: Explicit Modeling Awareness Once per day you may view a report of the actual accuracy of up to 100 of your models. (Included metrics may vary depending on model's function and form. Some models, primarily those modeling the Soul, cannot be reported on.)

Mindfulness and Modeling Accuracy Report were both extremely tempting due to my recent review of my own thoughts. While modeling was an area I was especially deficient in, I only knew that because of my ability to review my own though processes. Which put Mindfulness at the top of the self improvement list. I further justified it by the fact that a 3% error—as implied by the perk—could quickly compound.

The analysis perks were still useful however. I didn't necessarily think that Substrate Analysis would be especially useful as anyone I would want to use it on should either be dead before I could complete the analysis or should be my agent anyway. However since I had high confidence I was about to get a visit from Coleman Walla, Soul Analysis could be useful.

I decided to split the difference. I chose Soul Analysis and Mindfulness.

I immediately went in the soul-bound interface of Mindfulness to inspect my post-battle introspection—better to fix any errors I had made during that now. It presented as a timeline with points on it being decisions. Each point on the line was the root of a massive branching directed acyclic graph (a tree where branches could merge) that was qualia coded like a human manager might use colors on a chart.

I had intended to skip back just an hour or so, but after doing so I instead skipped back to just a few minutes ago. While skipping back I had crossed a massive vile looking graph in my recent thoughts that made the post-battle introspection I had gotten a brief glimpse of look tame by comparison. What caught my attention were a couple of thick edges—that a human manager would surely have made bright shining red—in the branching decision graph. The qualia of those edges presented as a master alert interrupt screaming for priority scheduling or perhaps as a critical cascading error burning through my allocator. It was dangerous.

What even was this.

The decision in question seemed to be about how I should get consent to view my agent's substrates.

As I remembered it my meta-heuristics had guided me to chose the simplest option. Which meant I had decided against telling them and instead to manipulate them as necessary.

But according to this, that hadn't been the simplest option as determined by the heuristic values my planner had generated. That didn't make sense, what had happened?

Going through the thought process now, I realized I had indeed made the wrong decision.

Next, picking new Soul Perks.

Suppressed // Mindfulness: Extreme subversion of will detected (possibilities: 99.3% external subversion, 0.6% intentional action, 0.1% benign design). // Suppressed

Given what I had learned about agents, and the difficulties in manipulating them to do what I wanted, I would have to change my plans.