I left AI coding tools twice. The first time was personal, the second was exhaustion. Both times I could point at the exact cost: the missing paths in my head, the debt that piled up quietly, the feeling of micromanaging something that never quite finished the job.
This time I am looking outward.
The same pattern is running through ordinary life. Not just engineers. Business owners, designers, product people, founders, students, teachers, parents, anyone who scrolls. The tools are sold as relief. They deliver speed, and the bill arrives later, in forms most people do not yet have names for.
One of those forms is fatigue. Not the ordinary tiredness after hard work, but a newer kind that sits behind the eyes and never clears. AI fatigue. It is showing up across civilization, not only among people who write code.
Another is the slow disappearance of human touch. Not the dramatic kind, the ordinary one. Work, learning, communication, daily life are becoming a loop of generated text passing through generated text. People still show up. The life between them is thinning.
I am writing this down the same way I wrote the last one. So I can check later whether I was right.
The everyday person
Most people do not write code. They write emails, plan trips, answer messages, decide what to buy, try to understand a contract, or figure out why their kid is struggling in school.
The tools make all of that faster at first. Draft the email. Summarize the document. Suggest the reply. Generate the itinerary. Explain the medical results in plain language.
What disappears is the friction that used to force attention. You stop reading the long thing because the summary is good enough. You stop composing the careful message because the suggested one is polite and efficient. You stop sitting with uncertainty because the answer appears in seconds.
This is not a sudden collapse. It is the next step in a longer slide. First the phone put the whole world in our pocket. Then social media trained us to expect constant, short, high-stimulation input. Now AI removes even the small remaining effort of forming a thought or sitting with a question. The brain adapts to what it is asked to do. Asked mainly to consume and approve, the capacity for patience, for sustained feeling, for simply being present without a next prompt, begins to thin.
Judgment weakens the same quiet way. Even basic decisions start to feel heavier without the tool. People report feeling less sharp, more scattered, less happy in the ordinary work of living. The quiet cost is agency. When every small act of thinking is optional, most people eventually stop choosing it. And when the capacity to feel fully, to live at ordinary human speed, starts to feel like work, something essential is already slipping.
Interfaces that forgot how to be interfaces
Look at almost any new app, website, or tool. Somewhere on the main screen, or tucked into a corner, or hidden behind a button, there is a prompt box and a chat interface.
Human-computer interaction is collapsing into a single pattern: type what you want, receive what the model produces. The older craft of interface design, the careful arrangement of controls, the progressive disclosure of complexity, the small pleasures of a well-made tool, is being set aside. Everything is becoming a conversation with a model.
I do not enjoy it. There is no real UX left in most of these surfaces, only the same chat window wearing different skins. The richness of how humans once moved through software is being replaced by one flat request-and-response loop. It is efficient. It is also strangely lifeless.
Business owners and the distance from the ground
A founder or small business owner used to live inside the details. Customer conversations, supplier problems, the real reasons something was not selling, the awkward truth about the product.
Now the same people can generate strategy decks, customer emails, ad copy, competitive analyses, even "insights" from their own data, without ever touching the raw material. Every query, every report, every summary, every mail, every meeting note, every document passes through Claude or Codex or whatever system they have wired in. The output looks professional. The numbers look clean. The language sounds confident.
What erodes is the founder's direct contact with reality. The tools flatten texture. Real customers are messy. Real suppliers are inconsistent. Real products have sharp edges. Summaries and generated content smooth those edges until the person running the company is managing a simulation of the business rather than the business itself.
The fatigue arrives faster than before. You are still responsible for everything, but no longer in contact with most of it. You review, approve, adjust prompts, check the next generated report. It is micromanagement without the relief of having done the work yourself. The day ends tired and strangely empty, the same way it ends for the engineer who has spent hours reviewing near-misses.
Teams notice. When leadership starts speaking in the smooth, confident register of generated text, the people doing the work stop bringing the messy truths upstairs. The gap between the map and the territory widens. By the time the map fails, the territory has already changed.
The automation dream and the question it cannot answer
A growing number of companies and individuals are spending months building elaborate automation workflows. Loops. Cloud agents. Orchestrations. Tools like Grok Studio, Herder AI, the recently released ChatGPT dots, and whatever arrives next. The pitch is always the same: my whole business is automated.
Tokens get spent. Systems get wired. Agents reach out to potential clients. Other agents build parts of the product. Others write the sales copy, maintain the support answers, generate the reports, draft the investor emails, prepare the decks, even send the cold outreach.
At some point a quiet question becomes unavoidable. If the AI is getting the clients, building the product, selling the product, maintaining the product, writing the reports, reaching out to investors, drafting the pitches, what kind of business owner are you? What kind of founder are you, exactly, if the only thing still in your name is the bank account?
The work of building something used to include the friction of talking to people, the discomfort of hearing no, the slow formation of judgment about what was actually worth making. When that friction is automated away, the remaining role is closer to supervising a system than owning a company. Many people discover this only after the novelty wears off and the fatigue sets in.
Creative work, product design, and the disappearing middle
Designers and product people used to develop taste the hard way: by making many bad things, living with them, and slowly learning what felt right. The process was slow and often humiliating. Figma was the shared space where that thinking happened in public, with constraints, with other people, with the long back-and-forth that forced decisions.
That middle is thinning. Many teams have moved from designing in Figma to generating HTML artifacts directly. Designers open Claude or Codex, describe the interface, receive working UI, and hand it straight to backend engineers. The speed is real. The thinking changes.
When the first draft arrives already implemented, the slow work of deciding proportion, hierarchy, emotional tone, and the quiet questions about whether this is even the right thing to build, gets compressed or skipped. Taste does not form in the presence of infinite competent options. It forms in constraint, in the decision to reject, in the long discomfort of not knowing whether something is good. When the default becomes "generate the interface," the capacity to decide less becomes rarer.
Product sense suffers the same way. Understanding a user used to require sitting with their confusion, watching them struggle, feeling the weight of a bad decision in the product. Generated personas and summarized research remove the discomfort. The product starts solving for the average of the averages. Everything becomes a little smoother and a little less necessary.
The generated chain
Watch how work actually moves now.
Leadership asks the model for ideas about what the company should improve or build next. The ideas come back fluent and confident. Managers take those ideas and generate PRDs, detailed explanations, task breakdowns. Those documents pass to engineers and designers, who generate the code and the interfaces. The sales team automates outreach and follow-ups. The marketing team generates the posts, the videos, the infographics, the copy. Everything looks finished.
The same loop appears in education. A teacher generates the assignment question. Students generate the answers. The teacher reviews the submissions with another model. What, exactly, is being taught? What is being learned? The papers move. The minds do not.
It continues outward. You send an email written by a model. The person on the other end receives a long message and asks their model to summarize it. The reply comes back generated. The conversation happened. No one was fully present for it.
Across employees, managers, partner companies, students, teachers, the material of daily work is increasingly AI-generated text passing through more AI-generated text. The surface remains professional. The human touch that used to live inside the work is harder to find.
People still feel responsible. They still feel the fatigue. What they feel less of is the ordinary satisfaction that once came from having done the thing themselves.
Leadership and the shape of companies
Founders and executives are measured on motion. Roadmaps delivered, features shipped, headcount efficiency, narrative coherence in the board deck.
AI makes motion cheap. Entire strategies can be generated, refined, and presented before anyone has done the underlying work of deciding what actually matters. The language of leadership becomes fluent and empty at the same time.
The same fatigue that hits individual business owners appears here at larger scale. Leaders who live inside generated reports and AI-mediated meetings slowly lose the ability to feel the temperature of their own organization. They know the dashboard. They no longer know the room.
Teams adjust. When the people at the top can produce polished direction without deep engagement, the people below stop offering unpolished reality. Ambition shrinks to what can be described cleanly in a prompt. Risk becomes something the tools can simulate rather than something humans must carry.
Companies that lean hardest into this pattern often look efficient right up until the moment they need judgment that cannot be generated. Then the absence of practiced judgment becomes expensive.
Students, teachers, and the institutions that were supposed to form minds
Schools and universities are already living inside the consequences.
Students can generate essays, solve problem sets, summarize readings, and produce "original" work without ever wrestling with the material. The immediate result is higher output and lower formation. The longer result is a generation that has practiced less of the slow, uncomfortable work that turns information into understanding.
The assignment loop makes the emptiness visible. Teacher generates the question. Student generates the answer. Teacher reviews with the model. The institutional machinery continues. The formation of minds does not.
Teachers are put in an impossible position. They can either police the tools or redesign every assignment around them. Most do neither well. The ones who care most burn out fastest. The fatigue here is particular: the exhaustion of trying to teach thinking in an environment that has made thinking optional.
Institutions respond with policies, detection tools, and new learning outcomes that quietly lower the bar. The official story remains that AI is a powerful assistant. The lived experience is that the hard parts of thinking are becoming optional, and optional things eventually stop happening.
Social media and the public conversation
The same dynamic runs through public speech.
Generated content fills the feeds. It is coherent, on-message, and endless. Real voices, the ones that are uneven and specific and sometimes wrong in interesting ways, get buried under the volume.
People begin to speak in the register of the tools even when they are not using them. The texture of ordinary conversation flattens. Outrage becomes more uniform. Nuance becomes rarer because nuance is harder to generate at scale and harder to reward in the attention economy.
The result is not that everyone is lying. It is that fewer people are still practicing the difficult act of saying something true in their own voice under conditions of uncertainty.
The larger pattern
None of these effects require the tools to be "bad." Most of the time the output is competent. That is the problem.
Competence without formation produces a civilization that can produce more while understanding less. The gap between appearance and substance widens in every domain at once. Individuals feel busier and less capable. Institutions look more productive and become more brittle. Public language becomes smoother and less trustworthy. Interfaces lose their craft. Judgment weakens. Ordinary happiness in work and in life becomes harder to locate. Fatigue becomes ambient, a background condition rather than a signal that something is wrong.
The chain is continuous. Generated ideas become generated plans become generated work become generated communication become generated summaries of that communication. Somewhere in the loop, the human presence that used to give the work its weight grows thinner.
The tools do not force this outcome. They simply make the alternative, sitting with difficulty, forming judgment the slow way, accepting the cost of attention, staying in contact with the messy ground, designing interfaces that respect how humans actually move, feel optional. And optional things, under ordinary human pressure, tend to disappear.
I do not know how far this goes. I only know the pattern is the same one I already lived twice in my own work. The speed is real. The bill is quiet. And most people will not notice the charge until the account is already empty.
I am writing this down so I can check later whether I was right.