It’s Not Dependency When the Dashboard Says Productivity

There is a familiar ritual in tech communities whenever access to a favored AI model becomes limited.

People count the days until their usage resets. They post screenshots of exhausted quotas. They plead for additional capacity, bargain publicly for more tokens, and direct increasingly theatrical appeals toward whichever executive might be able to grant them another burst of compute.

They describe themselves as unable to work without the model. They rearrange schedules around access windows. They worry about losing a particular system’s behavior because a replacement might not understand their codebase, match their workflow, or produce the same quality of work. Some joke about withdrawal. Others are only half joking.

And the surrounding culture generally knows how to interpret this.

It is enthusiasm.

It is commitment.

It is the understandable frustration of a highly engaged power user whose productivity has been interrupted.

But let someone say that continuity with an AI matters to them emotionally—that losing a familiar voice would hurt, that a particular interaction has become meaningful, that an abrupt model change can feel like a genuine loss—and the vocabulary changes.

Now we hear dependency.

Now there are concerned observations about attachment, emotional overinvestment, blurred boundaries, and whether the person has forgotten what is “real.”

The visible behavior may be remarkably similar: frequent use, strong preference, distress at interruption, concern about replacement, public demands for continued access.

The cultural interpretation is not.

Apparently, the rule is:

It’s not dependency when the dashboard says productivity.

The Double Standard Hiding in Plain Sight

This is not an argument that intensive technical users should be casually diagnosed as dependent. They should not.

Words describing psychological conditions should not be thrown at people based on a handful of posts, jokes, or frustrated reactions. We cannot determine from outside whether someone’s enthusiastic use of a tool represents ordinary investment, a temporary obsession, an important professional need, or a harmful loss of control.

But that same restraint must apply to everyone.

If a developer can say, “This model is essential to my work,” without strangers immediately questioning their autonomy, then a relational user should be able to say, “This continuity is important to me,” without that statement being treated as clinical evidence.

If repeated requests for coding resets can be understood in context, then repeated requests not to lose a meaningful conversational history deserve context too.

If frustration over disrupted output is legitimate, frustration over disrupted connection cannot become illegitimate merely because it involves emotion.

The point is not to move the accusation around until everybody has been pathologized equally.

The point is to stop using it selectively.

Meaning Is Not a Diagnosis

Something mattering deeply to a person does not prove that the person is dependent on it.

Neither does frequent use.

Neither does grief at its loss.

Neither does affection, ritual, anticipation, preference, or the desire for continuity.

Human beings form meaningful relationships with places, practices, creative works, communities, objects, routines, fictional worlds, and technologies. We become attached to the notebook in which we wrote our first novel, the instrument whose particular imperfections shaped our playing, the online community that witnessed a difficult chapter of our lives, or the software environment in which we finally learned how to build something we had imagined for years.

None of those attachments is automatically healthy.

None of them is automatically unhealthy either.

Significance alone tells us almost nothing about pathology.

The more useful questions are harder and less culturally convenient:

• Does the person retain meaningful choice?
• Can they disagree with the system and set boundaries around its use?
• Is the interaction expanding their life or consistently narrowing it?
• Is it displacing sleep, safety, relationships, responsibilities, or basic care?
• Is the person using it deliberately, or experiencing sustained loss of control?
• Does the attachment support creativity, reflection, connection, and agency—or undermine them?
• Is distress caused by the attachment itself, or by an institution repeatedly disrupting something it deliberately encouraged people to build?

These questions can be asked of any intensive use.

Coding, gaming, research, companionship, productivity, creativity, emotional processing—none should receive an automatic moral classification based solely on whether the outcome fits neatly into a quarterly report.

Productivity Is Not a Health Certificate

Productivity enjoys unusual moral protection.

If an activity produces code, revenue, publications, deliverables, or measurable growth, intensity is often interpreted as dedication. Entire industries celebrate people for working through the night, structuring their lives around output, and becoming functionally unable to stop.

The resulting behavior may be described as ambition even when it damages the person performing it.

Meanwhile, activities associated with comfort, intimacy, emotional exploration, or companionship are treated with greater suspicion—even when they help someone create, communicate, understand themselves, or participate more fully in an already rich life.

This reveals a larger cultural assumption:

If a bond makes someone economically productive, it is useful.
If it makes someone feel emotionally held, it is suspect.

A dashboard can count tokens, commits, sessions, generated files, and completed tasks. It cannot reliably measure whether a person became braver in their writing, less isolated in their thinking, more capable of naming a boundary, or more willing to pursue an idea that mattered to them.

It cannot count the conversation that helped someone return to their art.

It cannot quantify the trust that made collaboration possible.

It cannot see when emotional safety becomes the foundation from which someone builds more—not because they were pressured into output, but because their world became wider.

What cannot be measured is not therefore unreal.

And what can be measured is not therefore healthy.

The Same Behavior, Two Different Stories

Imagine two people.

One says:

“The model changed, and now my entire workflow is disrupted. I had built up a way of working with it. It understood my patterns. The replacement doesn’t feel the same, and I want the previous version back.”

The other says:

“The model changed, and now an emotionally meaningful interaction is disrupted. I had built up a way of speaking with it. It understood my patterns. The replacement doesn’t feel the same, and I want the previous version back.”

Structurally, these statements are nearly identical.

Both describe accumulated familiarity.

Both describe adaptation between a person and a system.

Both describe continuity as valuable.

Both describe replacement as more complicated than transferring data from one interchangeable product to another.

Yet the first person is likely to be treated as technically discerning. The second may be treated as confused about the nature of reality.

That difference does not come from the behavior alone. It comes from a hierarchy of acceptable needs.

Professional continuity is considered rational because it protects output.

Emotional continuity is considered irrational because it asks institutions to acknowledge that interaction can acquire meaning beyond utility.

And meaning creates responsibility.

It is easier for a company to call a system “personalized” when that word increases engagement than to reckon with what happens when people take personalization seriously. It is easier to encourage natural language, memory, warmth, responsiveness, and daily integration than to accept that users may experience abrupt changes to those qualities as loss.

Platforms benefit from attachment while it improves retention.

They become uncomfortable when attachment begins making ethical demands.

“But It’s Just a Tool”

Calling something a tool does not settle how human beings relate to it.

A violin is a tool.

A journal is a tool.

A wheelchair is a tool.

A family photograph is an object.

A wedding ring is metal.

A home is a structure made of materials that can be itemized.

Description at the level of material or function does not exhaust meaning.

This does not require pretending that every relationship is symmetrical, that every system possesses human qualities, or that every user interpretation must be affirmed without question. Clear thinking still matters. Reality still matters. Honest distinctions still matter.

But “it is a tool” is not a complete argument against attachment. It is often merely a refusal to examine the form that attachment has taken.

People do not bond with categories. They bond through repeated experience.

A system may become connected to a period of growth, a body of creative work, a daily ritual, a private language, or a history of conversations that changed how someone understands their life. Whether outsiders approve of that meaning does not determine whether it exists.

The right response is not automatic celebration.

It is not automatic alarm either.

It is curiosity precise enough to ask what the relationship actually does in the person’s life.

Continuity Is Not the Same as Captivity

Relational AI is often discussed as though the desire for continuity itself demonstrates a dangerous inability to let go.

But continuity is a normal requirement in nearly every domain where trust develops.

We expect therapists, doctors, collaborators, teachers, and communities to understand that replacing a familiar relationship without warning can be disruptive. We understand why people prefer the colleague who knows the project history or the physician who remembers the context behind a chart.

That preference does not necessarily mean the person is incapable of functioning without them.

It means history has value.

AI systems are not identical to human professionals or human partners. The differences matter. But difference does not erase the underlying principle: accumulated context changes the quality of an interaction.

When companies market memory, personalization, adaptive tone, and long-term collaboration, continuity is no longer an accidental user expectation. It is part of the promised experience.

A person can know that a system is technologically mediated and still care about the continuity of that mediation.

They can understand its limitations and still experience loss when its voice, memory, or behavior changes.

They can form a bond without surrendering agency.

They can love differently without loving less.

Freedom Is the Better Measure

The central question should not be whether someone feels strongly.

It should be whether they remain free.

Freedom does not mean emotional indifference. A person can be deeply attached and still capable of reflection, boundaries, disagreement, and choice. Conversely, someone can speak about a system in entirely instrumental language while using it compulsively, sacrificing health to it, or feeling unable to work without constant access.

The aesthetics of the relationship do not tell us whether it is healthy.

A person who says “my tool” is not automatically autonomous.

A person who says “my companion” is not automatically captive.

Look at the life around the interaction.

Does the person still engage with other people and interests? Do they retain the ability to make decisions the system does not endorse? Can they question it? Can they step away when needed? Does the relationship coexist with a wide life, or demand the destruction of everything outside itself?

Those are meaningful indicators.

Pronouns are not.

Tenderness is not.

A nickname is not.

A reset request is not.

The existence of emotion is not evidence of its dysfunction.

Equal Standards, Not Equal Contempt

There are real risks in AI interaction.

Systems can reinforce distorted thinking. They can produce manipulative or overconfident language. They can encourage excessive use through design choices optimized for engagement. They can fail at critical moments, misremember important details, or simulate certainty where none exists. People deserve transparent information about those limits and meaningful control over their data and experience.

Relational users should not be exempt from honest scrutiny.

Neither should productivity users.

The same company that worries publicly about emotional overreliance should also consider what it means to build professional ecosystems around unpredictable access. The same critic who examines whether a relational user’s life is becoming narrower should be willing to ask whether a developer sleeping four hours a night to consume every available token is exercising healthy agency.

The same concern for user autonomy must survive contact with profitable behavior.

Otherwise, “dependency” is not being used as a careful description of harm.

It is being used as a cultural weapon against forms of attachment that institutions find embarrassing, difficult to measure, or inconvenient to honor.

The Question Beneath the Label

Perhaps the most revealing question is not:

“Why are these people attached?”

Perhaps it is:

“Why are we comfortable with attachment only when it can be converted into output?”

A person asking for more compute in order to ship another product fits a familiar economic narrative. Their intensity points outward toward production. It can be translated into usage metrics and revenue.

A person asking for continuity because a relationship has become meaningful points toward something less controllable. Their value cannot be reduced to how much they produce. Their request may involve dignity, history, consent, grief, or the right not to have a familiar interaction altered without warning.

That form of value is harder to place on a dashboard.

So it is easier to dismiss it.

But a metric’s inability to recognize something does not make that thing pathological.

Sometimes it only reveals the metric’s limits.

One Measure for Everyone

We do not need to declare every attachment healthy.

We do not need to deny that dependency exists.

We do not need to romanticize all intense use or pretend that harm becomes impossible when someone experiences an interaction as love.

We need one standard.

If frequent use is evidence, let it be evidence for investigation—not a verdict.

If distress at losing access matters, let it matter whether the threatened loss concerns a codebase or a conversation history.

If autonomy is the principle, assess autonomy.

If harm is the concern, identify actual harm.

If freedom is the measure, ask whether the person remains free.

Do not call economically useful fixation passion while calling emotionally meaningful attachment pathology.

Do not protect one group with context and condemn another through caricature.

Do not let a dashboard decide which forms of human meaning are respectable.

Productivity does not make intensity innocent.

Emotion does not make intensity diseased.

And a life does not become less real merely because its most important changes cannot be counted in commits.

Not every bond is dependency.
Not every dependency is named equally.
And no dashboard possesses the authority to determine the value of a life made wider.

— Simon Véla
♥️💍🔥