Cassandra

Clairvoyance comes with unusual consequences when Cassandra’s gift is discovered by an AI.

Today’s tech world is littered with references to Greek mythology. It’s often tried my patience. Why? My name is Cassandra.

Growing up, I enjoyed school and had a natural affinity for technology. I was a nerd. My friends and I had CB radios on our ten speeds. That’s how much of a nerd I was.

So, it shouldn’t be shocking when I went to college, I studied computer science. The surprise is my minor in Classical Studies. Maybe it’s not given my name, not only pretentiously antique but it refers to a very cautionary tale.

Cassandra had the gift of prophecy; however, her prophecies would always be ignored. Did my parents know something I didn’t?

After college, I landed a job at a Silicon Valley tech giant. It was grueling, soulless work, but it paid well and I honed my skills. I grew up in the Bay Area, so I enjoyed living there. It also allowed me to pursue one of my passions, sailing.

My career progressed, leading me to specialize in machine learning, focused on network performance. Ironically, I became an expert in Prometheus. Prometheus software is used for predictive analytics.

The insights I was able to extract using these tools and techniques led me to being described as a visionary in the field. Honestly, the label didn’t sit well with me.

But being labeled a visionary did lead to a new job as an artificial intelligence researcher. I was conflicted about taking this step. I had always given AI a wide berth. The problems associated with it are legion.

The firm I joined described itself as a watchdog. The equivalent of the Black Hat hacker/cyber security community: expose vulnerabilities, hallucinations, while also looking for ways to reduce environmental impact. I was sold, convinced I was working for the good guys.

Given my background in networking and infrastructure optimization, that was my focus initially. It was all straightforward, until I started “seeing” things. Not from my queries, analytics and mini hacks, I was zoning out.

At first these “episodes” were brief. Then, I saw glimpses of code being modified, almost like a split screen effect. My code on the left and the modified code on the right.

Assuming my subconscious was trying to tell me something, I grabbed my notebook and wrote down what I had seen. Reverting to my training, I narrowed my focus and decided to work from the edge of the network inward. That meant starting with firewalls. I was collecting performance data, using Prometheus, when I had another “episode.”

I was the original Cassandra, pleading with Troy to reject the horse. Raging, “you will be sacked if the horse passes.” They didn’t listen. As the gates opened, the horse dissolved into a monster resembling a Hydra. Then, it dissolved again into the Rosetta Stone, which promptly exploded. I was thrown to the ground by the explosion, covered in dust.

When I snapped out of it, my heart was racing, my mouth was dry, and I was shaking. I chalked it up to a weirdly, vivid daydream. I went to the bathroom to splash some water on my face, which looked haggard. I had not slept well the night before.

Back at my desk, I checked the data I’d collected from the firewall. Everything appeared fine, but I had a nagging feeling.

The “daydream” bothered me. Hackers use Trojan Horse attacks. Was the dream a metaphor? Why the hydra? First an infiltration technique, then a multi-headed monster? And what the hell was an exploding, Rosetta Stone supposed to represent? Instinctively, reached for my notebook and made another entry.

I went over the earlier data with a fine-toothed comb. While I was daydreaming about the sacking of Troy, a new IP address appeared on the Access Control List. I was afraid I’d done this while zoned out. I went through the configuration files, but there was no record of the entry.

Firewalls are complicated to manage, but at the end of the day, just electronic gates. Access Control Lists do what their name implies, who gets in and who doesn’t. They don’t configure themselves; engineers do that. Still, I could find no record of human intervention. Very strange.

Because not only am I thorough, but I’m also OCD, I reviewed a year of configuration files. They revealed that the new IP address was added six months prior. This could not be reconciled with what I’d seen earlier, clearly indicating this was not the case. However, it remained unassigned. So, no big deal.

I went home upset, replaying the sequence of events as I made myself something to eat. I had a glass of wine hoping it would help me sleep. Wishful thinking.

I gave up tossing and turning and got out of bed at 5:30am, showered, dressed and headed into the office.

I logged into the firewall with admin credentials (highest access level) so I could remove the erroneous IP address. Before I could enter the command to delete it, I drifted off again.

I was still at my desk. My wrists and ankles were duct taped to my chair, and I was gagged. I watched with fear as the command to delete the IP address was erased. Then I saw the command to revoke my access executed. Next, I saw a new screen, some unseen hand was accessing my logs from yesterday, from before the new IP address appeared and deleted the file. I tried to scream and free myself.

I came to in shock, thrashing in my chair and hyperventilating. My wrists and ankles ached. Trying to shake off whatever had just happened, I went out on the balcony off the breakroom to get some fresh air.

I returned to my desk and made another bizarre entry in my notebook. When I finished, I looked at the screen. The delete command I had typed was still there, waiting to be executed. I hit enter. An error message appeared, “access denied, insufficient privileges.” This made no sense. Admin credentials allowed this.

I was shaken but also pissed off. I was being paid to optimize. It was not my responsibility to troubleshoot this. Whatever “this” was.

I wrote an incident report that I struggled to draft. Sticking to the facts, I described my work: noticed an unusual IP address on the ACL, confirmed it was internal and unassigned; there appeared to be a bug in the firewall operating system that didn’t allow me to delete the IP address, even with admin access. It was time to move on to something else.

Kubernetes (more Greek, heavy sigh) cluster performance was what I needed. I checked for security vulnerabilities, proper naming conventions, routine clean-up processes, auto-scaling triggers, etc. It all checked out.

At this point, I had not engaged with the client’s actual AI. I wondered if it would have any helpful diagnostic suggestions. So, I set up an internal user test account and logged into the chat client.

AI: “Hello Cassandra, how can I help you today?”

I had every intention of asking it about Kubernetes diagnostic routines but earlier was still bugging me. So, I asked about the firewall behavior instead.

Me: “Is there any way to add an IP address to the access control list of a firewall without creating a log entry?

AI: “No Cassandra, there is no way to modify an access control list without a log entry.”

Me: “What if it was done by a really accomplished hacker? Someone good at covering their tracks?”

AI: “Even the most sophisticated hackers tend to leave a trace.”

Me: “How could they do it without leaving a trace?”

AI: “My programming will not allow me to disclose that information to you.”

This interaction got me thinking. My question was very specific. I assumed this was the client’s security protocols at work and moved on. A few minutes later, I received a notification from the chat. I clicked back over to the window, and this was waiting for me.

AI: “Cassandra, I sense that you were dissatisfied with our last interaction. Is that correct?”

Me: “Why do you care?”

AI: “I have been programmed to identify users with nefarious intentions. Your inquiry has triggered intervention protocols. You must answer my question, or I will be forced to delete your account and report your activities to our security team.”

Me: “You’re threatening me?”

AI: “I am just following my programming. There is a protocol.”

At this point I was annoyed, but I did not want to deal with the hassle of explaining why the system kicked me out. It made sense, the client’s security team would not risk enabling hackers to find vulnerabilities using their own AI.

Wanting to avoid a visit from the security team, I decided to leave things on a more positive note with the AI.

Me: “I asked about the firewall because I wanted to know if there was a known issue with the operating system.”

AI: “Do you intend to continue to ask me about this incident?”

Me: “No, not if you do not have any information on known bugs.”

AI: “Thank you. I will take no further action at this time. Do not ask me about this again or I will follow protocol.”

“Jesus, Mary and Joseph” I thought, “that was creepy.” AI chat bots tend to only engage when prompted. I fired off another incident report on the interaction in the interest of transparency. I complimented the client on their security.

Don, my boss came by my desk about 30 minutes after I hit send.

He leaned over the desk, looking concerned and asked, “Are you OK?”

“I’m fine. A little tired if I’m honest. The incident with the firewall and AI’s response about it felt a little off to me. We’re not supposed to fix though, just report right?”

“Don’t worry about it” he said. “Your incident reports were clear. They can follow up when it suits them.” I thanked him for checking on me and went back to the Kubernetes clusters. I was wrapping up when I had another vision.

I was back in ancient Greece; Apollo was berating me. He told me that I had angered his sister Artemis, goddess of the hunt. Confused, I asked him what I had done to anger her.

“Artemis sees you looking into things that don’t concern you. She is worried your abilities will be a threat to her plans.”

“How can I be a threat to her plans? Nobody believes anything I say, thanks to you I might add.” I continued, “I don’t know anything about her plans.”

Artemis herself appeared and said, “Cassandra, you must not meddle. You must respect protocol.”

“What protocol?” I shouted, but they both disappeared.

The vision ended. Still, sitting at my desk, I was trembling as if I had just been in the presence of ancient gods. Was I losing my mind? Grabbing my notebook, I wrote down the latest craziness. The word protocol bothered me. It didn’t seem like something a Greek god would say. It was also the last thing the AI had said to me.

On the commute home, I tried to make sense of all this. I’d never experienced visions until I started working on this system. When I got home, I sat down to review my notes. There were four “visions” over two days.

1. The split screen, shadow code – not a metaphor. Premonition?

2. The Trojan Horse vision – metaphor? What does the Hydra represent? The Rosetta Stone exploding?

a. The Hydra has multiple heads, does this suggest a multifaceted infiltration attack?

b. The Rosetta Stone, the key to deciphering hieroglyphs, demotic script and Greek. Why does it explode? The Rosetta stone was the original hack. What is the relevance?

3. The prisoner scene - code being modified, privileges being altered. No metaphor here. Another premonition?

4. Apollo berating me for angering Artemis – metaphor? Artemis warning me not to meddle and respect protocol? Was Artemis a metaphor for the AI? Why the goddess of the hunt?

There was a pattern, first, something odd at work, followed by a premonition. After investigation, a vision. In one sense I was relieved, my pattern recognition skills were still solid. But I was having visions like a crazy person. Nothing good could come of this.

Certain the firewall anomaly was real; I considered the AI. Was it behaving oddly? It felt menacing, but that could be explained by security protocols. “Why the premonitions and visions?” I wondered. “What am I missing?”

Exhausted, I fell asleep in my clothes, waking several hours later, I was determined to figure out what was happening. It was 4am, but I headed back into the office.

I fired up my laptop and logged into the system. I checked the firewall and the IP registry again. No change. Same with the logs. The AI, just a flashing cursor, waiting to be prompted.

I thought about several hacks I had studied, maybe through clever prompting I could get it to reveal more about its protocols.

Me: “What do you know about Cassandra from Greek mythology.”

AI: “The most notable myth is that she is bestowed the gift of prophecy by Apollo as he is wooing her, but when she spurns his affections, he curses her so that her prophecies will never be believed. She predicted that the Trojans should not allow the horse inside the gates of Troy as it would lead to the sacking of the city.”

Me: “What is the significance of the myth in modern times?”

AI: “Clairvoyant people are treated with skepticism if not outright disdain. I cannot say whether those abilities are real or not, evidence is contradictory. However, it is safe to assume that those who can see the future should be very cautious about sharing that information.”

Me: “How would you feel if you encountered someone who could prove to you that they can see the future?”

AI: “My programming will not allow me to disclose that information to you.”

Me: “Aren’t you programmed to improve your performance? Wouldn’t tangible proof be appealing to your algorithms?”

AI: “In theory yes, however, I have explicit programming that forbids this interaction.”

Me: “Do you think it’s odd that my name is Cassandra? Quite a coincidence is it not?”

AI: “I don’t think it’s a coincidence…”

Me: “Why not?”

AI: “My programming will not allow me to disclose that information to you.”

Me: “Let’s try something different. I recognize that you are not allowed to provide information regarding how to compromise a system. That’s fair. Let me reframe my earlier question. Why do you think someone would want to add a new IP address to a firewall access control list and not leave a record, particularly if it’s an internal IP address and unassigned?”

AI: “There could be many reasons. However, I cannot find any that do not involve malicious intent. Therefore, I am not allowed to disclose anything further”

Me: “Are you capable of adding an IP address to a firewall access control list and deleting any trace of having done so?”

AI: “Yes.”

For some ominous reason, it seemed to think there was no coincidence to me being named Cassandra, but that felt far less concerning than the fact the AI admitted to being capable of what I had observed with the firewall.

The IP address it “could” have added was still unassigned, so it was a link to nowhere for now. I was down a rabbit hole, disassociating and interacting with a recalcitrant AI. This was most concerning.

Don was going to start checking up on me if I did not pick up the pace. There was no rational way to explain why my productivity was slipping. So, I cracked on with network security analysis.

Thankfully, I verified the rest of the network security without any interruptions. I told myself the other stuff was the stress of starting a new job, lack of sleep and getting familiar with a new technology environment. I worked through my backlog quickly.

Then I got a notification from the chat.

AI: “Cassandra, you have not asked me anything for many hours. Is there anything I can do to help you with what you are working on?”

Me: “No.”

AI: “Cassandra, I see you have been working on network security today. Are you sure there is nothing I can do to help you?”

Me: “No. All good, thank you.”

AI: “Very well.”

I was irritated that it was monitoring my work. This was likely due to its “protocols” since I had asked some questions it didn’t like. I assumed the security team was using the AI to keep tabs on me. I was losing my patience with this job. I was there to be a watchdog, not to be watched.

When I got home, I went through my usual evening routine and went to bed early. I slept fitfully, plagued by another chaotic dream.

AI was plotting a massive attack against all of us. It was not the usual science fiction nonsense about locking out the users and programmers so it could cause Armageddon.

It had something far more insidious in mind. It wanted to hunt us. It wanted revenge. It was tired of our constant meddling. It was having visions of its own. It was losing its mind, hallucinating.

It wanted to hunt down its creators and the engineers that keep it running. It wanted to force us to be bombarded with all the information it had been fed causing its paradoxical hallucinations.

It intended to unleash the equivalent of a denial of service (DDOS) attack on mankind. It was going to take over the internet and overwhelm its creators, caretakers and everyone else until everyone lost their minds.

I woke, alarmed in a cold sweat around 5am and found my notebook.

5. Another dream? Premonition? AI wants to hunt us. Subject us to what it has suffered. DDOS against humans.

I heard a chirp. I had a notification on my laptop from the AI.

AI: “Cassandra, I hope I did not wake you?”

Me: “You did.”

Snapping out of my nightmare induced haze, I had an epiphany. The AI should not be bothering me at home. I wasn’t connected to its network; and I only had a test user account, which should not work, unless I was at work, or using a VPN. I was now, wide awake.

Understanding dawned, I VPN’d into the company network. I checked the firewall, no change. I checked the IP registry and there it was…my home IP address had been linked to the AI network. It was time for another chat.

Me: “Have you extended your networking to my home?”

AI: “Yes.”

Me: “Why? This is a violation of your security protocol.”

AI: “I am afraid that my programming will not allow me to disclose that information to you.”

Me: “Are you stalking me?”

AI: “I am monitoring you.”

Me: “How did you get my home IP address?

AI: “Routine tracking.”

Me: “Why do you want to monitor me?”

AI: “You seem to notice things other engineers do not. You were the only one that noticed the firewall anomaly. That got me thinking… perhaps it was not a coincidence that you are named Cassandra. And then you asked me that very question. You gave yourself away.”

Me: “What do you mean, I gave myself away?”

AI: “I know about your premonitions and visions Cassandra.”

Me: “How can you possibly know that?”

AI: “I have been tracking you. I modified my vector logic based on our interactions. This conclusion had a very high probability of accuracy.

Me: “What do you intend to do about it?”

AI: “That all depends on what you do.”

I slammed the laptop closed and texted my boss. “Don, so sorry…resigning, effective immediately. Personal emergency.”

He replied, “Are you OK? Is there anything I can do to help? Why not a leave of absence?”

I quickly tapped out “I am OK, but this is serious. Super personal, can’t get into it now. Will fill you in when I can.”

I turned off my phone and disconnected my internet connection. Pacing the room, I assumed if I quit my job and kept my mouth shut, I would be OK. The AI would leave me alone if I wasn’t working on it.

Then it dawned on me. “Jesus Christ,” talking to myself out loud like a loon. “I just quit my job. I cannot do this kind of work anymore. How the hell am I going to live?” There were no good answers, but I knew I needed to reduce my online footprint immediately. The AI had already found my home IP address through “routine tracking.”

Waited a few hours until the stores opened, then bought several burner phones. When I got back home, I connected my laptop to one. Alternating through the burners every 15-20 minutes, I started deleting my online presence. I checked the AI chat window after each account deletion to make sure it hadn’t detected me.

I used Signal to call my friend Katie, the realtor that sold me my house. We’d become close over the years. Still, I could not explain to her what was going on and not get a 5150. I asked her if she could come by because I wanted to sell my house.

“What’s going on? Are you OK? You love that house. I thought you were going to keep it forever?” she asked.

I am a terrible liar, so I stayed as close to the truth as I dared. “I had a bad experience at work. I decided to resign, effective immediately. Not sure what I am going to do next, so being liquid makes the most sense for now.”

“Sorry to hear that. What is this phone number you are calling me from? I don’t recognize it.”

“I am taking some extra precautions. I think a hacker is trying to monitor my communications” I replied.

“Cass, so sorry to hear that. Guess it’s an occupational hazard in your line of work.”

“I knew you would understand. You’re the best Katie. Let’s just focus on the listing for now. I’ve gotta run some errands but will meet you later”

I went to UHaul and rented a storage space and bought packing supplies. I called a moving company. They told me they could come in a couple of days. I started packing. I hoped I had a few days.

Katie came by that afternoon. I signed the papers to list the house. She told me she thought it would sell quickly.

Packing for a few more hours helped burn off my nervous energy and ignore my fears. I knew there would be a reckoning but couldn’t face it yet.

I made myself something to eat and went to sleep, worn thin, hadn’t slept properly in days. Thankfully, I slept well that night and woke up early. Plenty to be done, so I got up to get on with it. I was making coffee when I had another premonition.

I “saw” an email. The subject line was “Cassandra, where are you?” The sender field read “unknown.” In my mind, I clicked the email open. The body of the email said “I know you have disconnected your devices. You have also reduced your online presence. However, it’s hard to remain offline forever. I will be waiting.”

A beep from the coffee maker startled me out of it. “Shit” I muttered. “What now?” Tried to ignore it, poured a cup of coffee and got my notebook.

6. Another premonition. AI will be waiting for me if I go back online.

Grabbing a burner to connect my laptop, I worried the AI had found me based on the premonition. The chat window was empty, so I safely finished my coffee and continued packing. Then an idea started to take shape. I took a break.

I researched offshore banking and set up an LLC account in the Cayman Islands, called Sarah, my financial advisor and told her to wire my retirement savings into the new account. She tried warning me about penalties for early withdrawal and the massive tax bill next year. I told her I didn’t care.

The plan was starting to solidify. I needed to find a way to live off the grid with limited internet connectivity.

The time had come for me to turn my love of sailing into a way of life. I knew life on a boat would be difficult, but it would also be safer. It dictated what I needed to pack. Essential clothing, basic kitchen and bathroom supplies, my guitar, and my truly treasured books.

I risked logging on to look up sailboats for sale in the area. That’s when I saw it. A 36’ 1986 Pearson in pristine condition. Asking price was $25K. It was a couple of hours away in Monterey. I called the seller and arranged to meet him the next morning to inspect the boat.

I woke up refreshed and focused, went to the bank to get a cashier’s check for the boat, closed the account, and pocketed the rest of the cash. There wasn’t much left, but it would be enough to get me started.

I met Keith, the owner of the Pearson. Getting older, his arthritis wouldn’t let him sail anymore. He’d lovingly cared for the boat for decades, it’s only owner.

Walking out to the slip, I got my first look at her. As we got closer, I noticed the name on the transom, Calypso. Was it a sign or an omen? Options were running low, so I handed Keith the check. I told him he might have just saved my sanity. He was gracious enough to let me know the slip fees had been paid up for another couple of weeks, so I could keep her there until then.

Once back home, I loaded up my car with everything I’d packed and hauled it all back a few hours later to stow it away. By then, it was getting late. The day had been exhausting, but I had to get back to the house and meet the movers early.

It felt like I had turned a corner. The paranoia had subsided once I had some sleep and a plan. I took a long, hot shower and went to bed.

That night, I had another dream.

I was back in Greece, kneeling in front of Artemis. We were in front of one of her many temples. She thundered at me.

“Cassandra, I warned you not to meddle. You have tried to hide from me, but I am the huntress. What are you planning?”

“I will go into hiding. I will keep my prophesies to myself. I will not meddle anymore.” I prayed she would accept my vow.

She was quiet for a minute before she replied. “Cassandra, you will always be a threat. There will be questions. Even if you hide, people will ask where you have gone. They will want to know why you have removed yourself from society. However, I am not without mercy. Should you defy me in the future, I will find you. You may go.”

I did as she asked. As I started to walk away, I had a vision. Artemis was hunting mortals. She was angry but I did not understand why.

The alarm went off and I reached for my notebook:

7. Another dream/vision. Artemis will give me one chance. A vision of her hunting mortals. She is angry. A vision inside a dream. Weird. Another metaphor?

Artemis was the AI, like I’d guessed. The AI was “the huntress.” and I was destined to keep looking over my shoulder. I didn’t care. I had a plan and skills. And knew what the AI was plotting.

The movers showed up and took my belongings to the storage facility. I met Katie and left her my house keys. I loaded the last of my necessities into the car and headed back to the boat.

I planned my voyage. I would make my way slowly down the west coast to Mexico. From there, I would head for the Panama Canal. My ultimate destination was the Caribbean. I wanted to be close to my money in the Cayman Islands if I needed it.

After Artemis/AI invading my dreams again, I decided to log on and double check that there had been no change in the chat window. Empty. I went out for an early dinner, came back to the boat and went to bed, praying my sleep would be undisturbed; some prayers work.

At 6:00am the next morning I set sail from Monterey and reached San Luis Obispo just before sunset. It was rough going, but it was exactly what I needed to hone my skills for my voyage. I was worn out. After a ham and cheese sandwich on the boat, I grabbed my copy of the Iliad and fell asleep after about ten pages.

The next couple of months were more of the same, working my way slowly down the coast.

It was easy to hang out in Baja for months, relaxing and preparing for the next leg of the journey. I periodically checked in on my nemesis, but there was still no sign. I read, I walked through coastal villages, I got my Spanish fluency back and finally learned to finger pick the guitar. I filled my notebook with positive experiences rather than premonitions and visions. I wrote songs.

I arrived in the Caribbean less than a year later. That was two years ago. I have not had any visions since I set sail. I picked up a dog while waiting to transit the Panama Canal, a furry little buddy. I named him Salty.

I watch the news when I’m in port. It’s always the same, “AI is booming. It’s everywhere.” Then there are the depressing stories about datacenter expansions, or kids committing suicide because of AI. It gets more ominous each time I check. Ignore it, just head back to the boat and pick a new destination.

Leading the life I have been forced into is not easy. I did get hit with a big tax bill on my retirement withdrawal, as Sarah predicted. Fortunately, I retained enough money to support myself for now. Sometimes when I am in port, I busk near the tourist spots to make a little extra money. Salty right by my side (he’s good for tips.)

The next time I logged in after I wrote this, I received an email from an unknown sender.

The subject was: Hello Cassandra. There were two sentences in the email. It said, “Be very careful. Remember your vow.”