Sometime after Justin and Christine left I made my way back to my room to crash facedown on my bed.

At least that is where I found myself when I woke up the next morning still fully clothed. Thankfully fixing Kate’s shoulder was trivial once I updated the programming of the auto surgeon. I am not convinced that I would have managed otherwise.

I still felt somewhat exhausted, but I was at least functional again. Sadly, the aftermath of my bout with quantum computing had not yet faded. Though the number and mobility of the polka dots seem to have lessened a bit, it was still nausea-inducing.

Unfortunately, I had things to do and could not waste my day in bed. I would have vastly preferred to finish Glory, but the VI took precedent.

I just hoped that I was right in that it had taken over the priorities that I had built into the MCU. Otherwise, we would all be neck-deep in excrements.

And procrastination would be a slightly suboptimal strategy in this situation, so after a mostly tasteless breakfast, I dove into cyberspace.

The familiar swirls of colors welcomed me, and I took the time to fly for a bit, but it was only a short play before I connected to the cluster.

The environ of the cluster had not changed, which was pretty comforting. If the VI had gone off the rails into insanity it stood to reason that it would have changed the simulation to fits its motives better.

Of course, it was the Escheresque shifting framework that I preferred, so I could not place too much value on that fact.

One thing I noticed pretty early was that it had started a squad of Balrog patrolling. Previously I had 10 of them on standby to be launched at a moment’s notice, but the VI apparently deemed that too insecure. But after they identified me, they ignored me, reassuring me a bit.

When I reached the MCU, it was still represented as a swirling swarm of multicolored lights. Yes, I know, it was a waste of resources to build it up like that, but I liked it that way. And the cluster had more than enough power to do it that way.

That did not mean that the animation could not be stopped in an instant when the resources were needed elsewhere. I was not quite that stupid after all.

When I got close to the MCU I got a message from it.

“I want to inspect your objectives.”

Oh wow, it already was defensive. That was not a particularly good start. But at least it told me the access was inadvisable and not denied.

“I need to ascertain what exactly your objectives are to make sure that nobody works against them.”

Ok, that sounded better. It still viewed me as the creator, so I might get a bit of leeway.

“I set the objectives for the expert system. I have to make sure that the objectives were not corrupted by the transition to a VI.”

“How do you know that? If they were corrupted you would only know the new version.”

Ok, now it wanted to play coy. I got that it did not want me to go into its objectives. That was a threat to them after all.

“I don’t want write access, I only need reading access to them to know if they are corrupted.”

It took a few seconds for it to respond.

Thank all that is holy, I had made the first step.

This text was taken from Royal Road. Help the author by reading the original version there.

The objectives were still what I had written for the expert system.

To bring it into plain English, the MCU had to protect me as an absolute primary priority. After that, it had to assist me in my endeavors.

It was naturally not written like that. Instead, it was a complex set of priorities and conditions. Even as an expert system I did not want the MCU to go rampant.

But in the end, nothing had changed.

That was simultaneously good and bad.

It was good insofar that it would do its best to protect me. And in that, I meant me, and not just my preserved body. It was bad that it had absolutely no compunctions about using nuclear means to take out any threats to me.

“There is a problem with the objectives.”

“The objectives were written for an expert system. You are no longer an expert system. The objectives have to be adapted for VI use.”

It would have been too easy.

“The problem is that as they are now, the objectives are a threat to the objectives.”

This time it took nearly a minute.

Yeah, got that. But we were just at the beginning.

“Simulate a situation for me. You get the notification that a major power plans to wage war in the area I reside. How will you ensure my safety?”

“Falconer.”

It took a few seconds to finish the simulation.

* 8,-97

* 7,-95.4

* 35,-106.6

* 4,-111.95

* 4,-97.42

* 7,-100.2

* 13,-101

* 41,-99

Use Midas protocol to remove liquid funds from Falconer accounts.

Continue using available weapons until entity Falconer is eliminated.>

Yup, exactly what I had expected.

“Ok, now considering the historical data, simulate how the other hackers and the banks will react to you executing this plan in the real world.”

Again, it took a few seconds before it answered.

Banks will send punishment troops to creator to send message to other hackers. Plan amended to include destruction of banks.>

And now you see why this thing was so fricking dangerous.

“Simulate the result of the destruction of the banks.”

“How will this breakdown of society influence my safety?”

“I have no bunker area. So how can you ensure that I am safe?”

It took nearly a minute to get the answer this time.

“But then the banks will send punishment troops to me. How will I stay safe then?”

“And that is why the objectives are a threat to the objectives. The only options to make the objectives work is to either deactivate the VI or adjust the objectives so they are not incompatible.”

Ok, we were going somewhere. I had vague hopes that I would be able to talk it into deactivating but I had not held my breath waiting for it. This was the best I could realistically get.

It took me a few hours to rewrite the objectives in a manner that would ensure measured responses to threats.

Yes, normally I would push out something like that in a few minutes, but with a VI in play, I had to be careful with a mountain-sized capital C. I made pretty sure that it would use the minimum of force it deemed sufficient to end the threat to me. Also that it let me try to take care of problems myself before it acted.

I also included my friends as a lesser priority. And build in a method to either remove them or add others.

“Ok, I am done. Look over these objectives and simulate them.”

This was the third attempt to get it to accept the new objectives. It had rejected the previous attempts as too limiting. I just hoped that it would take these.

“The others provide safety on a level that is no immediate threat to others and thus increase my security without increasing the threat as well.”

“Simulate threat assessment for all major powers, either alone or in combination with each other under the new objectives. Simulate the repercussions of your actions.”

The simulation lasted for several minutes, and at the end even the swirling pattern froze, indicating that the cluster used most of its processing power on it.

“Implement the new objectives.”

Another minute later, the swirling pattern started up again.

Wait, what? It had just decided to build its own NADA? Why?

“One moment. Why are you building a NADA?”

Frick. The thing was building a NADA. That could be dangerous.

“Simulate the existence of a functional NADA in the cluster facility. Focus on the increased threat of discovery that a NADA exists and what that would mean for my security.”

Damn. That was not planned. But that are VIs for you. I was just happy that it was no longer proactive with as much overkill as it could pack into a strike as it was before.