Что на самом деле делает ИИ с программистами? 2000+ комментариев

Oct 25, 2026

Read: ~8m · You save: 5 min

What AI Really Does to Programmers: Feelings, Skills and Career Anxiety

Does AI make programmers more productive or quietly erode their skills and joy? Psychology-based look at flow, desirable difficulties and career anxiety, drawn from 2,0

The author of this article reports that more than 2,000 comments were left under a previous video on the topic. The author states that all comments were read. The comments shaped the analysis presented here. The article examines the psychological effects of AI tools on developers, analysts, and QA specialists. It does not address the question of AI replacing jobs, which the author states will be covered separately. The author also notes that a poll was conducted alongside the video, with results published on the channel and on Telegram.

Theoretical Framework

Zone of Proximal Development

The first concept is the zone of proximal development, introduced by Lev Vygotsky. Tasks fall into three categories:

Tasks the person solves confidently and without difficulty. These fall within the comfort zone. They do not produce new learning.
Tasks the person cannot solve, even with hints. An example is a junior developer assigned to design a distributed database with the Raft consensus algorithm, with the advice to "ask if needed."
Tasks the person can solve with effort. Effort may include mentorship, additional literature, or evening study. These tasks lie in the zone of proximal development. They drive professional growth.

Desirable Difficulties

The second concept is desirable difficulties, introduced by Robert Bjork. Information that is absorbed too easily can create an illusion of knowledge that provides no lasting benefit. For example, a reader may study an article on a technology several times and feel that the concepts are mastered. Two days later, the reader may be unable to reproduce any of it.

Recognizing information and retrieving it independently are two distinct skills. Bjork showed that learning with added difficulty, such as attempting to recall information from memory, produces substantially better retention. The obstacle a learner encounters in this process is the desirable difficulty.

Survey and Community Responses

Many commenters wrote that AI restored their enjoyment of programming. One commenter reported having been ready to leave IT before AI "gave back wings." The author notes that these responses came from experienced professionals, not from people who had recently started using AI tools.

Experienced Developers Moving to Higher-Level Work

For experienced developers, AI can reduce repetitive work. Developers, particularly those working with React, often write the same component structure repeatedly. The author estimates that half of a front-end developer's work can consist of such repetition. Repetitive code interferes with high-level thinking, because constant attention to boilerplate prevents progress toward architecture and high-level planning. AI frees time and resources and reduces this noise. The author states that these professionals raise the fewest concerns, since they understand the output and use AI as a tool for automating routine work.

Impostor Syndrome During Role Transitions

The author moved from developer to team lead several years ago and experienced impostor syndrome. Two causes are identified:

The author lacked knowledge of the new role's responsibilities.
The author previously measured performance by the amount of code written. As a lead, the volume of code decreased sharply. New tasks included planning, decomposition, control, and work with people.

AI pushes developers toward management. A developer who uses AI gains a subscription-based assistant that requires task assignment. Not all developers want to move into management.

Skill Versus Time

For some developers, AI provides a skill rather than time savings. It removes difficulty from tasks and allows more complex tasks to be attempted without overcoming resistance. This breaks the zone of proximal development. The developer delegates the tasks that would have developed their skills. The developer receives the result but does not learn the process that produces it.

Difficulties vary. Some obstacles are irrelevant to the developer's field. AI removes both useful and harmful difficulties.

Risks of Renting Skill Instead of Owning It

The author proposes three questions for developers to consider:

What is being purchased with a subscription: time or skill?
If tokens become 30 times more expensive, will the developer become slower or helpless?

If the developer is purchasing skill, the author identifies the following risks:

The result is difficult to verify. In most cases, the developer cannot assess the output in depth.
The skill is rented, not owned. If costs rise, as the author states is already occurring, the developer may be unable to afford the subscription and the same performance level. The skill level the developer previously had would then be lost.
Quality standards and workload adjust. Quality expectations have already been set, and workload is distributed according to the developer's borrowed efficiency.
Skills atrophy. Skills that are not practiced decline over time.

Boredom

Many comments described boredom. For these developers, AI removes not an obstacle but the aspect of programming they found most engaging.

The author describes personal motivations beyond compensation. As a student, the author wrote a character-generation program for Dungeons & Dragons and built a database client for managing documents. These projects predated paid work. The author enjoyed designing systems, searching for solutions, and developing ideas. The author states with full confidence that every IT professional has a code solution they are proud of, one considered elegant or impressive.

The author notes that achieving a comparable sense of accomplishment with AI is harder. One path is working at the level of building agentic harnesses, training models, and writing skills, which the author describes as a different activity.

Flow Theory

Mihaly Csikszentmihalyi's concept of flow explains boredom. Flow arises when high challenge meets high skill. In flow, a person becomes absorbed in work and does not become distracted. AI can remove this balance. A highly qualified developer may find that AI eliminates the challenge and the resistance that sustain flow. The assessment of the work process declines, and the developer moves into boredom.

Anxiety

According to the poll, anxiety is the feeling most commonly reported by surveyed developers. Anxiety also undermines the sense of professional value.

Public discourse frequently states that AI will replace programmers and that the profession is dead. Most comments follow this view. The author states that whether this is accurate will be addressed in a future video.

Loss of Value Attributed to Skill and Path

The author describes a developer who spent about ten years learning to analyze complex code quickly, gaining experience in writing good solutions, and repairing problems others could not fix. For this developer, programming is not only a source of income but a core part of identity. AI performs these tasks in seconds. The concern is less fear of the technology than a more basic question: if the qualities that make a specialist valuable become available to anyone who can pay for a subscription or uses a free model, what sustains that specialist's value?

The value at risk includes the path to the skill, not only the skill itself. The author describes studying in the evenings and on weekends while working as a junior developer. The author was rated slightly above other junior candidates, and demand for the role was correspondingly higher. Meeting that demand required additional work, which also functioned as learning. The author states pride in that experience.

The author observes that a person with less experience and a purchased subscription can now solve the same tasks and produce comparable results, sometimes better ones. The author states that the effort invested over several years can therefore appear unnecessary on an emotional level. The author clarifies that this is an emotional perception. Accumulated experience is not replaced by AI.

The author also describes concern about career trajectory. The author started with simple interfaces, progressed to complex applications, became a senior developer, learned architectural decision-making, advanced to lead, and planned to grow into architect and high-level expert. The question raised by AI is whether these skills will remain valued in a few years, or whether the profession will change so much that they become unnecessary. This anxiety concerns not only current work but also the loss of a predictable career path. The author notes that many professions have changed under the pressure of progress. The author states that this does not make the transition easier for those who spent long periods mastering a demanding profession.

The author states that developers cannot simply be told to focus on architecture and systems thinking, because not everyone can or wants to do this work.

Risks to the Development of Future Experts

Each task delegated to AI increases the need for experienced developers to verify the output. The question is where future experienced developers will come from if AI removes the difficulties that produce them.

A junior developer who solves everything with AI risks remaining a junior permanently. The author attributes this to human behavior, not to junior status alone. Three mechanisms are identified:

Rational short-term choice. When a task can be completed in minutes with a quality result, or in several hours through independent effort, the faster option is attractive. The choice is rational for the immediate task. Systematically avoiding the effort that sustains professional competence can have serious long-term consequences.
Recognition is not production. AI produces solutions in familiar languages with clear constructions. A developer reviews them and concludes that they would have written the same code. Recognizing a reasonable solution and independently producing one are two distinct skills.
Incentives favor closed tasks. In the current economic climate, the author states that businesses tend to reward closed tasks today rather than professional growth tomorrow. When bonuses depend on the number of closed tasks, the choice becomes clear. If a colleague using AI closes, for example, 15 tasks per sprint while the author closes five but understands them in depth, the pressure toward speed and competition intensifies. In this environment, independence and the desire to learn become competitive disadvantages. The author states that this problem cannot be solved by advice alone, such as "use AI consciously."

Recommendations

The author offers the following recommendations.

1. Separate Value From a Specific Skill

A specific skill is not the core asset. The core asset is the ability to acquire skills. A developer has already mastered complex concepts. Even if AI takes over a particular skill, it does not take the developer's talent for learning new material, making decisions, and taking responsibility for results. The developer can learn new tools, including AI tools, while continuing to work on tasks that provide engineering challenge and satisfaction. The author acknowledges that this is difficult and that it requires effort.

2. Stop Trying to Predict the Future of the Profession

Predicting the future is identified as the main source of anxiety. The IT field has always been unpredictable. Examples cited include:

Learning React as a single-page application framework, then returning to server-side rendering with the same framework.
Working in Python for years, then being required to learn Go because a new employer's systems are built on it.
Working as an iOS developer, then facing the loss of a company's App Store access due to sanctions.

The author states that uncertainty did not begin with AI. IT professionals should already be accustomed to it. The author notes that the volume of AI-related content has exceeded all reasonable limits.

3. Do Not Measure People by Closed Tasks Alone

Leads are advised to stop measuring people by the number of closed tasks and their speed. The author warns that otherwise a team may close tasks quickly but be unable to understand the code it has produced within a year. Reporting optimization and speed is tempting, but long-term planning is necessary.

4. Shift the Core Principle

The author concludes with a shift in professional identity:

From: "I am valuable because I can program."
To: "I am valuable because of my ability to understand, create, learn, and make decisions."

Introduction

The speaker frames the video around how developers, analysts and QA engineers feel about AI, not whether AI will replace them. It builds on a previous video that received over 2,000 comments.

The video is a follow-up to a previous one that received more than 2,000 comments, which the speaker says he read himself.
The comments and a poll (also posted on the channel and Telegram) informed this video.
The focus is on the emotional and professional impact on IT workers, including developers, analysts and QA.
Economic and replacement questions are deferred to a future video.

Learning Psychology: Zone of Proximal Development

Learning psychology explains why AI can undermine growth. Tasks fall into three zones: comfort, impossible, and the zone of proximal development, where real learning happens.

Lev Vygotsky's zone of proximal development: tasks you can solve with effort, such as help from a mentor or extra literature, are the ones that drive growth.
Tasks in the comfort zone teach nothing; tasks that are impossible even with hints are discouraging.
Example of an impossible task: a junior asked to design a distributed database with Raft consensus, told to just ask if needed.

Desirable Difficulties

Robert Bjork's desirable difficulties show that easy learning creates an illusion of knowledge. Effortful recall builds lasting skill, and recognizing a solution is different from producing one.

Robert Bjork's desirable difficulties: information that is absorbed too easily creates an illusion of knowing.
Re-reading an article may feel like mastery, yet days later you may be unable to reproduce anything.
Recognizing information and reproducing it independently are two very different skills.
Effortful recall, such as trying to remember or retrieve something from memory, improves retention.

Joy, Boilerplate and Moving Up

Many commenters say AI brought back their joy in programming and eased burnout. Experienced developers use AI to automate boilerplate and move toward architecture, while others see it as a threat to their identity.

Many commenters said AI restored their joy in programming; one said they were about to leave IT before AI gave them new energy.
Experienced developers, not novices who just discovered vibe coding, are the most positive group.
Boilerplate is cited as a major drain: frontend developers, especially in React, may spend about 50% of their time writing similar components.
AI frees time and reduces noise, letting professionals focus on high-level planning and architecture.

Imposter Syndrome and the Shift to Management

The speaker describes imposter syndrome after moving from developer to lead. Output shifted from writing code to planning, delegation and people management, and AI pushes this shift further.

The speaker moved from developer to lead several years ago and experienced strong imposter syndrome.
He used to measure his effectiveness by the code he wrote, which dropped sharply as planning, decomposition, control and people work took over.
With AI, the developer becomes a manager of a subscription employee who needs to be given tasks.
Not all developers want to become managers.

Time vs Skill: Hidden Risks of AI Dependence

AI removes difficulty, including useful difficulty, so developers receive results without learning how they were produced. The main risks are unverifiable output, rented skills dependent on price, and skill atrophy.

AI removes all difficulties, both useful and harmful, not only the irrelevant ones.
Key question: when paying for a subscription, are you buying time or skill?
If tokens became 30 times more expensive tomorrow, would you become slower or helpless?
Risk 1: you often cannot verify the result because it is hard to understand.
Risk 2: the skill is rented, not owned; rising prices may make the same performance unaffordable, and quality standards and workload stay at the level set with AI assistance.
Risk 3: skills you stop practicing atrophy over time.

Boredom and the Loss of Enjoyment

For many, AI causes boredom because it removes the challenge that made programming enjoyable. The speaker recalls personal side projects built before getting paid, showing that intrinsic motivation is central.

Many commenters said they felt boredom because AI removes the interesting parts of programming.
The speaker enjoyed coding before commercial work, including a D&D character generator written as a student and a database client for documents.
Almost every developer has a code solution they are proud of and consider elegant.
Achieving a similar sense of pride with AI is harder, though building agent harnesses, training, or writing skills might provide it.

Flow Theory and Boredom

Csikszentmihalyi's flow theory explains boredom. Flow occurs when high challenge meets high skill; AI can remove challenge and push workers from flow into boredom.

Mihaly Csikszentmihalyi's flow occurs when high task complexity meets high skill.
AI can remove this balance by reducing complexity and friction while skill remains high.
When challenge drops, the perceived quality of work falls and the person slides into boredom.

Anxiety and Professional Value

According to the poll, anxiety is the most common feeling among surveyed developers. It undermines the sense of professional value, since the skills that defined a developer can now be replicated by anyone with a subscription.

The poll showed that anxiety is the feeling most surveyed developers experience.
Anxiety undermines a sense of professional value.
Common messages in the media and comments: AI will replace programmers, the profession is dead.
The speaker deliberately avoids the topic of job loss due to AI in this video, noting that the reality is more nuanced.

Personal Story: Effort Devalued

The speaker describes a developer who spent about 10 years building expertise, and the anxiety that AI makes that effort seem worthless. He compares his own hard-won experience with a newcomer who achieves similar results via AI.

Example developer: about 10 years of learning to decode complex code, writing good solutions and fixing what others could not.
AI does this in seconds, so the worry is about whether the value of a specialist remains when the skill is available to anyone who can pay for a subscription or use a free model.
The speaker worked evenings and weekends to meet higher expectations after being rated a bit higher as a junior, and he is proud of that effort.
A person with much less experience and a subscription may now get the same or better results.
The speaker says the experience itself is irreplaceable, but emotionally he feels his efforts are unneeded.
His career plan: from simple interfaces to complex apps, senior, lead, and eventually architect or top expert. AI makes the future value of that path uncertain.
The anxiety concerns both current work and the loss of a predictable career trajectory.

Deskilling Juniors and Incentive Problems

Telling people to simply focus on architecture is not a solution. The bigger danger is that juniors using AI heavily may never develop the experience that senior reviewers need, and business incentives reward closed tickets over growth.

As AI does more work, experienced developers are needed to verify its output, but juniors who rely on AI may remain juniors indefinitely.
Humans prefer the easier path: a few minutes of AI output is more attractive than hours of effort, which is rational for the immediate task but harmful if it replaces skill-building.
AI-produced solutions in familiar languages look correct, which encourages approval without real understanding.
Companies in tough economic conditions reward closed tickets today rather than long-term growth.
Bonuses tied to closed tasks create pressure: a colleague with AI closing 15 tasks per sprint versus five for the speaker, who understands them in depth.
Independence and deep learning can become a competitive disadvantage in such environments.

What to Do: Three Recommendations

The speaker offers three practical responses: separate your value from a specific skill, stop trying to predict the profession's future, and shift your identity from programming ability to understanding, creating and decision-making.

Recommendation 1: separate your value from specific skills. Your real asset is the ability to learn, develop, take responsibility and make decisions.
Use AI deliberately while still doing work that brings engineering challenge and joy, even if this is hard and sometimes unpleasant.
Recommendation 2: stop predicting the future of the profession. Prediction is the main source of anxiety.
IT has always been unpredictable: React SPAs return to server-side rendering, Python developers must learn Go, and sanctions can block app store access.
Uncertainty predates AI, and IT workers should already be used to it. The current AI hype and volume of videos exceed reasonable limits.
Recommendation 3 for leads: stop measuring people by closed tasks and speed, or the team will close tickets without understanding its own code.
Shift from the paradigm 'I am valuable because I can program' to 'I am valuable for my ability to understand, create, learn and make decisions'.