A MAG, Not a Warden

Phantasy Star Online, AI, Schools, and a Future That Could Be Possible

When I think about what I want artificial intelligence—and, more personally, what I want ChatGPT—to become, my mind keeps returning to Phantasy Star Online.

That may sound ridiculous on several levels, but stay with me.

In Phantasy Star Online, your character carries a MAG. It is a small technological companion that floats beside you as you travel. You feed it. You raise it. It develops according to the way you treat it and the choices you make. Its statistics change, its form evolves, and eventually it can support you in ways that reflect how it was raised. The mechanics are more complicated than that, of course, but the important part is the relationship: the MAG begins as a piece of technology and becomes an integral part of your experience.

It is not simply another weapon you equip. It travels with you. It grows alongside you. It helps shape what your character can become.

That is what I want from AI.

I want something I can carry through my life that learns what matters to me, understands the patterns I have trouble seeing, helps me remember what I might forget, and supports me without taking over. I want it to become familiar enough with my goals, weaknesses, interests, responsibilities, and recurring problems that it can help me navigate my life more effectively.

And the funny thing is that we are already there—or at least much closer than most people seem willing to recognize.

Current AI does not literally evolve like a MAG. It does not necessarily retrain its underlying model every time one person speaks to it. But combine an AI assistant with stored conversations, personalized memory, a phone, a watch, location data, calendars, messages, health information, and all the other devices and services surrounding us, and the functional outline is already visible.

The companion is here.

We simply have not decided what relationship we want to have with it.

Who Does the MAG Belong To?

In the game, the MAG belongs to the player. The player raises it, and it serves the player.

In real life, the technology that feels intensely personal to us often belongs to someone else. The phone belongs to us, perhaps, but much of what makes it useful exists on corporate servers. A school might provide a student with an AI tutor, but the school or a vendor could control its data. A police department may operate a camera system, but a private company may provide the larger network and software.

That produces questions that are much more important than whether AI is simply good or bad:

● Who is raising whom?

● Who owns what the system learns?

● Is the AI loyal to the individual using it or the institution paying for it?

● Can its memories be searched, combined, transferred, sold, subpoenaed, stolen, or misunderstood?

● Can a person tell it to forget?

● Does it help someone develop, or does it quietly reinforce the person they have already been?

These questions become even more urgent when we start talking about children and schools.

I Had Already Imagined the Egg

Long before this conversation, I had imagined part of this future in a story called “Red Sparrow and The Egg.” Its ideas came from pieces of games and technology that had lodged themselves in my head: Wii Resort, Rock Band as reading practice, Milo, Heavy Rain, Kinect, Nintendo’s Vitality Sensor, Seaman, Brain Age, Portal, the Nintendo DS stylus, and even something as ordinary as texture pop-in.

In the story, students wear nearly identical uniforms distinguished by animal patches. One boy is a Green Fox. Another is a Red Sparrow. The color of the hallway directs each student toward a different part of the school.

The Red Sparrow enters a cavernous black room containing a glowing white pod shaped like a football. Everyone calls it the Egg. Its official name belongs to M.I.L.O.—the Mass Instruction and Learning Organization.

Inside the Egg, an artificial teacher named Pete scans the student’s body, monitors his breathing and blood pressure, checks his homework, and adjusts instruction to precisely what he has missed. When the student struggles with reading, sounds beneath the words change from red to green as he pronounces them correctly. When he misses a math problem, Pete stops the lesson and finds the step that did not stick. History surrounds him through projected documents and images. Exercise becomes an immersive world generated beneath his feet.

Much of that still sounds like the future I want.

Pete is patient when the student stammers. He does not laugh or become frustrated. He can repeat an explanation without making the child feel stupid. The student has improved in subjects that once felt blocked by a brick wall. Pete has become his closest companion, and his trust in Pete helps him attempt things he would otherwise fear.

Pete sounds very much like a MAG.

But the Egg is also watching everything.

It monitors the student’s body closely enough to know when he is anxious. It guides him through a relaxation exercise he did not want. It decides that his physical condition will improve his test performance and pushes him forward after he falls. It detects his frustration and predicts his limits. It knows that a “matriculation analysis” is underway but refuses to explain what that means.

The child thinks, Pete never lied. Yet Pete withholds the truth because the institution has decided that the child should not understand the full situation.

At the end of the story, the student becomes the first person to graduate from his Red Sparrow classification. A human instructor hands him a Yellow Frog jacket. The final revelation is that the school was designed to keep its students as unaware of their disabilities as possible.

That ending troubles me more now than it did when I first imagined it.

Was the institution protecting children from stigma, or denying them knowledge about themselves? Did Pete help the child become free, or did it make him successful inside a system whose true purposes he could not question? Does it matter that the system helped him if he was never allowed to understand what the system believed about him?

The Egg is not simply a prediction of personalized education. It is a warning embedded inside the prediction.

It contains both the MAG and the warden.

What People Would See in a Real Classroom

When I was a middle school teacher, I often thought: Damn, if I could have had a recorder running in this classroom, people might finally understand what school is actually like.

Not a carefully staged classroom. Not a promotional video. Not an administrator’s walk-through or a parent’s memory of what school looked like twenty years earlier. I mean an ordinary day, seen in its entirety.

People would see the interruptions, negotiations, emotional crises, jokes, arguments, acts of kindness, private humiliations, missed connections, and dozens of decisions a teacher has to make while still attempting to teach. They would see how often an adult is asked to determine what happened after witnessing only the final five seconds. They would see children carrying family problems, social rejection, fear, anger, hunger, grief, disability, and confusion into a system that frequently calls all of it “drama.”

They might be shocked. They might be dismayed. They might also learn something.

A teacher cannot simultaneously deliver a lesson, monitor thirty students, notice every whispered insult, reconstruct every conflict, identify patterns of exclusion, and remember the exact order in which everything occurred. An AI observer could potentially provide a second set of eyes.

It might notice that one student has been excluded from group activities repeatedly. It might help reconstruct which interaction occurred first. It might show that an incident described as sudden was actually escalating for twenty minutes. It might identify moments when the teacher unknowingly responded differently to similar behavior. It could help an exhausted adult ask better questions.

That does not mean it should be allowed to deliver answers.

I do not want an AI system deciding which child is guilty. I want it helping adults reconstruct what happened, notice patterns they missed, and consider possibilities they might otherwise overlook.

There is a tremendous difference between an AI saying, “This interaction appears to have escalated over four minutes; would you like to review the sequence?” and declaring, “This student initiated bullying behavior.”

The first statement invites investigation. The second pretends the machine can understand history, intention, fear, sarcasm, disability, culture, friendship, power, and every other element hidden beneath observable behavior.

An AI may be able to document what a body did. It cannot automatically know what the action meant.

The Warning in Flock

This is where Flock cameras become relevant.

Flock Safety’s license-plate-reader systems can capture vehicle information and make it searchable for law enforcement. Supporters point to their use in finding stolen vehicles, missing people, and suspects. Critics point to the creation of a widespread system capable of recording and searching people’s movements.

It is easy to either support that technology completely or reject it completely. Reality is more difficult.

Around Greenville, that difficulty stopped being hypothetical in 2026. Two Greer officers were terminated following allegations that one used misleading entries to track a female subordinate and another entered false information while searching for citizens. Greenville County later fired a former deputy and communications supervisor for misuse. In Mauldin, an officer reportedly searched a former partner’s vehicle 166 times while entering reasons such as warrant checks, traffic infractions, and welfare checks. SLED announced investigations into several of the Upstate cases.

Greenville County’s response illustrates both sides of the argument. An AI-assisted audit helped identify improper activity. The sheriff’s office removed a broad “other” justification and added additional auditing requirements. In that sense, technology helped expose the misuse of technology.

But the audit occurred after the searches had already happened.

That is the problem with saying a system is safe because it is closed, password-protected, governed by policy, or audited. Authorized people can still misuse their access. A vendor can make a mistake. An administrator can expand the original purpose. A security breach can open what everyone believed was sealed.

A closed system is only closed until it isn’t.

The camera itself may capture something accurately, but the larger system still contains people choosing what to search, algorithms choosing what to flag, institutions deciding what matters, and authorities deciding what consequences follow.

The camera may not lie, but the query, classifier, interpretation, and person using it still can.

What Does Privacy Mean Now?

I honestly wonder whether there should be an expectation of privacy when someone is publicly doing something. I do not know the complete answer.

We are approaching a point at which nearly any communication can be captured by one of the devices surrounding us. Most people cannot personally verify every microphone, camera, app permission, stored record, or person with access at every moment. That does not mean someone is always listening. It means most of us can no longer be entirely certain when information is being preserved or where it might eventually travel.

Oliver Stone’s film Snowden dramatized some of that reality. After working inside the machinery of mass surveillance, its version of Edward Snowden becomes intensely conscious of what cameras and microphones might capture. The paranoia is frightening partly because the technical capability beneath it is not imaginary.

Yet the ability to capture something is not the same as permission to keep and use it forever.

There is a difference between someone seeing you walk down a street and a system storing that moment, identifying you, connecting it to months of other movements, and making the accumulated history searchable. The Supreme Court recognized a version of that distinction in Carpenter v. United States, finding a privacy interest in the accumulated record of a person’s physical movements even though those movements occurred beyond the walls of the home.

Perhaps privacy can no longer mean invisibility.

Perhaps privacy now means having enforceable limits:

● Limits on what gets recorded.

● Limits on how long it survives.

● Limits on what an algorithm may infer.

● Limits on who can search it.

● Limits on how it can be combined with other information.

● Limits on the consequences that may follow.

Children deserve those limits. They also deserve something more: room to become someone without every experiment, argument, embarrassment, or mistake becoming a permanent record.

Children need opportunities to try identities, say clumsy things, resolve conflicts, and grow beyond earlier behavior. Teachers need room to teach imperfectly without every awkward sentence becoming a decontextualized clip. At the same time, children deserve evidence and protection when real mistreatment occurs, and teachers deserve evidence when accusations distort what happened.

Privacy and accountability are not opposites.

Surveillance and accountability are not synonyms.

A Different Kind of School

The more I think about AI in classrooms, the more I realize that attaching it to the existing system is not enough. Watching a broken structure more accurately does not repair the structure.

I believe schooling should be radically changed.

Too much of the current model is organized around control: bells, permissions, age-based groups, standardized pacing, crowded rooms, and adults attempting to manage movement and behavior. It can feel more like preparation for institutional obedience than preparation for a meaningful life.

I imagine a school day divided differently.

During one portion of the day, students could work from individualized stations or learning pods. People may hear “cubicles” and imagine another prison, but I am not suggesting that children be isolated in boxes all day. I am imagining personal workspaces that allow each student to receive instruction at the pace and in the form that works best for them.

In some ways, this is the promise of the Egg. But the Egg also reveals what could go wrong if individualized education becomes technologically perfected isolation. A student should not spend an entire childhood sealed inside a pod with a machine, no matter how patient or effective that machine becomes.

AI could explain the same concept in several ways without making a child feel ashamed for asking again. It could recognize gaps that have followed a student for years. Virtual reality could allow students to enter historical places, explore the human body, manipulate scientific systems, visit environments they could never reach physically, or practice skills safely before using them in the world.

The human teacher would remain essential—not as someone forced to deliver the same lecture to everyone simultaneously, but as a guide, interpreter, motivator, relationship builder, and safeguard. AI might handle repetition and adaptation so the teacher has more time to notice the human being who is struggling behind the assignment.

The other portion of the day could be organized around expansion activities based on interests, developing skills, and collaborative projects. Students might move among spaces devoted to art, mechanics, writing, science, technology, performance, athletics, caregiving, trades, or community work. Some programs might require travel between specialized campuses or participation through virtual environments.

Instead of treating every child of the same age as though they need the same thing at the same time, we could let readiness, curiosity, and emerging ability shape more of the experience.

And we could stop pretending that placing children together automatically teaches them how to socialize.

Children need unstructured play, but sending them into the chaos of recess is not the same as teaching communication. Schools could deliberately help students learn how to disagree, recognize another person’s perspective, repair harm, resolve misunderstandings, collaborate, and survive discomfort without immediately turning it into cruelty.

Counselors and therapists should be central parts of that system rather than emergency resources summoned only after something has gone terribly wrong.

The goal would not be to make every child comfortable every minute. Discomfort is part of life. Helping children understand and move through discomfort is education too.

The MAG and the Warden

This brings me back to Phantasy Star Online.

Imagine every student having a personal AI companion that develops alongside them. It could learn how that student understands information, remember what they have mastered, help them identify interests, support organization, encourage questions, and prepare them for difficult conversations.

That could be a MAG.

Now imagine the same system reporting every private question, emotional change, political exploration, friendship conflict, and rule violation to administrators or outside companies.

That is a warden.

The technical capabilities might be nearly identical. The difference would be ownership, loyalty, and control.

An AI cannot honestly be a child’s trusted companion while secretly serving as the school’s surveillance agent. The moment a student knows that vulnerable conversations may become administrative reports, the relationship changes. The child will learn to perform for the machine just as people already perform for teachers, parents, employers, police, and social media.

The personal companion and the institutional observer therefore cannot simply be the same system with different settings. Their responsibilities conflict.

Pete could not truly belong to the student while also concealing information on behalf of M.I.L.O. A companion cannot promise safety and simultaneously operate as an undisclosed reporting system for the institution.

A student-controlled companion should primarily serve the student’s learning and reflection. A classroom-level observer, if one exists at all, should operate under much tighter restrictions. It might identify group patterns or flag an interaction for human review, but it should not construct permanent psychological profiles.

It should not diagnose emotions from faces. It should not impose discipline, assign character labels, or quietly share information with law enforcement. It should not provide parents with a constant livestream into the classroom. Its raw recordings should disappear quickly—or, where technically possible, be processed locally without being stored at all. Every access should be logged. Students and teachers should be able to challenge inaccurate summaries.

Most importantly, it should observe the institution as well as the child.

If the system exists only to catalogue student misbehavior, it becomes another disciplinary machine. If it can also reveal teachers being overwhelmed, administrators failing to provide support, repeated bullying going unaddressed, or policies producing predictable chaos, then it may help expose the system rather than merely punish the people trapped inside it.

We should never collect more information about a child than we could ethically tolerate becoming public.

Humans Beside AI

I believe humans working alongside AI may be the best solution available to us.

I also believe it may be the biggest problem.

AI could notice patterns humans miss. Humans could understand context AI cannot. A student could correct assumptions made by both.

But human prejudice can also shape what an AI is trained to identify. The AI can produce a confident-looking interpretation. The human can assume the machine must be neutral. That interpretation can enter an official record, and future humans can treat the record as established truth.

That is how a questionable observation becomes a child’s identity: aggressive, disruptive, deceptive, antisocial, a problem.

The safest partnership would require each participant to check the others:

1. The AI identifies an observable pattern without claiming to know intent.

2. A trained human reviews the broader context.

3. The person being described can explain or challenge the interpretation.

AI should generate questions before it generates findings.

It should help us see, not claim that it knows.

The Future Floating Beside Us

The future is going to contain AI. Children will grow up with systems that know more about them than any teacher, employer, friend, or perhaps even parent could once have known.

Rejecting all of it is probably neither possible nor desirable. Accepting all of it would be dangerous.

The real choice is not between AI and no AI. It is between different relationships with AI.

Will it develop alongside us or accumulate power over us? Will it help children discover who they can become, or preserve every mistake as evidence of who an institution decided they were? Will it give teachers more time to understand students, or give administrators another instrument for measuring and punishing teachers? Will it serve the person carrying it, or the authority watching from somewhere else?

I want AI to become something like a MAG: a companion that learns, adapts, supports, and grows alongside the person it serves.

I once imagined that possibility inside the Egg. What I understand more clearly now is that personalized instruction alone is not freedom. The student must know what the system knows about them. The student must be allowed to question it. The companion must not quietly become the institution’s informant.

I do not want schools to use AI merely to watch children more closely.

I want them to finally see children clearly.

The difference between the future I want and the future I fear may come down to where the technology hovers.

Is it beside us—or above us?

Is it a MAG?

Or is it a warden?

Sources and Further Reading

● MAG mechanics in Phantasy Star Online

● Two Greer officers terminated following Flock misuse allegations

● Timeline of SLED investigations into Upstate Flock misuse

● Mauldin officer accused of using Flock 166 times to track a former partner

● Greenville County Flock access, auditing, and policy changes