AI now often arrives already installed. It appears inside email clients, document editors, spreadsheets, design tools, and code environments, not as a separate presence but as part of the application itself. The assistance is immediate and context aware, shaped by what the app assumes you are trying to do.
This embedded AI is convenient. In Outlook it proposes replies before the tone has settled. In Word it reshapes sentences toward clarity and efficiency. In Excel it infers intent from columns and formulas. In code editors it completes functions before the problem has fully surfaced. Each instance is helpful within its own frame, yet each remains bound to the logic of the tool that contains it.
Over time, this binding begins to matter. Thinking stays inside the application because the AI does. The AI augments the app, but it does not escape it. What emerges is not a general thinking partner, but a collection of function specific assistants, each fluent only in the language of its host.
It is often at this point that some people step outside. Text is copied. A plain text editor opens. The act is simple, almost mundane, yet the effect is immediate. Outside the app, AI becomes app agnostic. The same AI that could assist in Outlook or Word can now respond to an email, a reflection, a philosophical question, or an unfinished thought without being told what kind of object it is supposed to produce.
This exit is not dramatic. It is quiet, reversible, and deeply intentional. It marks a shift from being assisted inside a structure to thinking in a space that has no structure at all.
App Bounded Thinking and Its Limits
Every application encodes a theory of work. An email client understands communication as correspondence that should be sent. A document editor understands writing as something that moves toward completion. A spreadsheet understands thought as tabular relations that can be calculated and summarized.
When AI is embedded inside these tools, it inherits their theories. It learns what counts as a task, what counts as progress, and what counts as success. The assistance it provides is therefore never neutral. It is shaped by the app’s definition of what the user is doing.
This shaping is subtle. There is no coercion, only suggestion. Yet suggestions accumulate. Questions are framed in ways the system can answer quickly. Ambiguity is resolved early. Language becomes efficient, sometimes at the expense of resonance or care.
Function specific AI excels at this kind of narrowing. It performs well precisely because the problem space is constrained. But the cost of this constraint is that thinking begins to adapt itself to the tool rather than the other way around.
The result is not shallow thinking, but bounded thinking. Certain questions are easier to ask than others. Certain forms of uncertainty feel out of place. Reflection that does not immediately point toward output starts to feel unproductive.
At some point, the discomfort is no longer about efficiency. It is about freedom.
Plain Text as an App Agnostic Space
A plain text space has no built in expectations. It does not know whether you are writing an email, drafting an essay, thinking through an ethical dilemma, or simply noticing something that does not yet have a name.
Because of this, it is fundamentally app agnostic. Nothing in the environment privileges one kind of outcome over another. Language appears without being classified.
This neutrality changes how thinking unfolds. A professional message can sit beside a fragment of reflection without friction. A half formed sentence does not need to justify its existence. Questions can remain open without triggering a suggestion engine.
In such a space, AI also behaves differently. No longer bound to a function specific role, it becomes a conversational partner. It can respond to tone, intention, hesitation, and revision without assuming what the final artifact should be.
The experience is often described in sensory terms. It feels quieter. Lighter. Less hurried. The mind moves differently when it is not being guided toward a predefined end.
This is why plain text feels like refuge. Not because it is protected from complexity, but because it allows complexity to remain unresolved for longer.
Copy and Paste as Discernment
In many technical circles, the elimination of copying and pasting is celebrated as progress. Removing this friction can indeed improve flow when the work is already well specified.
But in reflective work, copying and pasting plays a different role. It is not merely mechanical. It is an act of selection.
To copy something is to decide that it matters. To paste it into a plain text space is to invite it into a context where it will be reconsidered. This pause creates a moment of discernment.
What appears inefficient from a productivity perspective functions as care from a cognitive one. The act slows the process just enough for judgment to surface. Context is not assumed, it is chosen.
When AI removes this threshold entirely, content flows freely, but attention does not always follow. The distinction between what was given and what was intentionally brought forward begins to blur.
In an app agnostic practice, this threshold remains visible. Each movement of text carries intention. Each inclusion is deliberate. The friction becomes a form of responsibility.
Dialogue Versus Agentic Action
As AI systems grow more capable, they increasingly offer to act rather than respond. Tasks can be delegated. Workflows can be executed autonomously. Decisions can be made in the background.
This shift toward agentic AI is often framed as liberation. Less effort, fewer interruptions, more time reclaimed. For certain kinds of work, this promise is real.
Yet for thinking that is exploratory, ethical, or relational, agentic action can feel like a loss rather than a gain. When AI acts on behalf of the user, the shared space of attention disappears.
Dialogue depends on turn taking. It depends on the ability to pause, question, redirect, and dwell. In a dialogical relationship, thinking remains visible. Responsibility stays close to the moment of formation.
Agentic AI collapses this space. The work is done, but the thinking has been skipped over. The human remains accountable for outcomes without having remained present for the process.
This is why some prefer AI that responds rather than acts. Not because they fear automation, but because they value partnership. They want the AI to stay with them, not move ahead of them.
Function Specific Power and Its Narrowness
Many advanced AI systems are impressive precisely because they are function specific. Some excel at writing code. Others at managing citations, formatting documents, or navigating formal academic workflows.
Within their domains, these systems perform remarkably well. But they assume that the problem has already been shaped. They expect the question to arrive in a recognizable form.
For thinkers whose work begins before formalization, this assumption can feel constraining. The system is ready to help, but only once uncertainty has been reduced to specification.
The disappointment that follows such encounters is not about capability. It is about misalignment. The system is optimized for finishing, while the thinker is still trying to understand what is worth beginning.
In contrast, an app agnostic plain text dialogue allows thought to remain fluid. It tolerates questions that do not yet know their destination. It allows imagination to wander before it commits.
This openness is not a lack of power. It is a different orientation toward thinking itself.
Staying with the Question
Choosing an app agnostic way of working is not a rejection of modern tools. It is a choice about where thinking should happen.
It reflects a desire to protect spaces where language is not immediately instrumentalized, where AI does not default to function specific behavior, and where agentic action does not replace dialogue.
In such spaces, questions are allowed to remain questions. Thought is allowed to mature at its own pace. Intelligence becomes something shared rather than deployed.
As AI continues to integrate more deeply into applications, the challenge will not be to keep up with its capabilities. It will be to notice where thinking still feels alive.
Plain text offers a reminder. Before there are apps, there is language. Before there are functions, there is attention. Before there is delegation, there is dialogue.
Staying with the question may not be the fastest path forward. But it remains one of the few places where meaning has room to appear.
Image: StockCake
Tom’s Blog 