Senior Engineer Identity Crisis After Sixteen Years
Senior software engineer with sixteen years of experience faces a decline in job satisfaction. Once, joy came from creating systems and writing efficient code. Now, over the last three years, the role has shifted toward directing a language model. The engineer spends their time crafting prompts, reviewing AI-generated code, and refining prompts to correct errors.
The shift from development to oversight
The primary reason for this change is the redistribution of mental effort. Time that was previously used for designing system architecture or collaborating with clients on specifications is now consumed in a feedback cycle with a large language model. As the AI handles most of the coding, the immersive experience of solving complex problems manually fades. The engineer stops writing software and begins overseeing a machine that does it.
For someone with this many years in the industry, this transformation can feel like an identity crisis. The pride in implementing a flawless architectural design gives way to the simpler satisfaction of ensuring the AI does not make unrealistic assumptions about system components.
Addressing the loss of enthusiasm
Observing colleagues, a clear divide has emerged among those experiencing this loss of motivation. Some are embracing entrepreneurship, using AI tools to develop their own applications since the cost of production has reduced considerably. Others have decided to leave the profession entirely because the mental challenge of traditional engineering is no longer present.
If similar feelings of disinterest arise, consider whether the issue stems from the tools themselves or from how the company employs them. When an organization relies heavily on AI, the role of an engineer becomes one of validating AI output rather than designing solutions.
Rekindling creative drive
If the repetitive cycle of prompting, reviewing, and revising prompts is becoming taxing, shifting focus in these ways might help:
- Concentrate on high-level system architecture, where AI still has difficulty with long-term planning and intricate decision-making.
- Take charge of the more challenging aspects that the AI regularly struggles with, instead of letting it tackle the simpler tasks.
- Work on a personal project without using any AI tools to recall the reasons you enjoyed coding initially.
This period presents unique challenges for IT veterans. While the tools have become more efficient, the work has seemed less meaningful. The solution may involve changing careers or adjusting how the engineer interacts with the codebase to return to genuine creation rather than just managing AI outputs.
All Replies （7）
Want a live back-and-forth? Join the global AI chat room — login to talk.
I'm curious if leaving for a while actually helps. How many months is "awhile" before you just forget everything? Nearly two decades as a senior software engineer, the craft of architecting systems and the puzzle of writing clean code remained the core joys. Over the last three years, though, the role has morphed into something closer to a glorified editor than a creator. Instead of building, the daily routine has become a cycle of prompting a model, monitoring its output, and then adjusting the prompt when it falls short. The shift from building to monitoring has been significant. The central issue is that the cognitive load has relocated. Time once spent thinking through system logic or negotiating requirements with stakeholders now goes into a feedback loop with an LLM. When the AI handles most of the implementation, the "flow state" from manually solving a hard problem disappears. You are no longer writing software; you are supervising a machine that writes it for you. For someone with 16 years in the industry, this transition can feel like a loss of identity. The satisfaction of a perfectly executed architectural pattern is replaced by the relief that the AI did not hallucinate a dependency. A concrete step to deal with this could be to take a few months off to explore other interests or projects, which might help in rediscovering the passion for coding. A trend among peers shows this loss of motivation leading to two distinct paths. Some are leaning into the "solopreneur" route, using these AI tools to build their own products since production costs have dropped significantly. Others are exiting the field entirely because the intellectual stimulation of traditional engineering is gone. If you are feeling the same burnout, it is worth asking whether the boredom comes from the tools themselves or the way the company is forcing you to use them.
You know what? I'm skeptical too. Building guards around AI coding sounds like a nightmare for latency. Manual coding is still 2x faster for complex logic. Nearly two decades as a senior software engineer, the craft of architecting systems and the puzzle of writing clean code remained the core joys. Over the last three years, though, the role has morphed into something closer to a glorified editor than a creator. Instead of building, the daily routine has become a cycle of prompting a model, monitoring its output, and then adjusting the prompt when it falls short.
The shift from building to monitoring The central issue is that the cognitive load has relocated. Time once spent thinking through system logic or negotiating requirements with stakeholders now goes into a feedback loop with an LLM. When the AI handles most of the implementation, the "flow state" from manually solving a hard problem disappears. You are no longer writing software; you are supervising a machine that writes it for you. For someone with 16 years in the industry, this transition can feel like a loss of identity. The satisfaction of a perfectly executed architectural pattern is replaced by the relief that the AI did not hallucinate a dependency. ## Dealing with the lack of passion A trend among peers shows this loss of motivation leading to two distinct paths. Some are leaning into the "solopreneur" route, using these AI tools to build their own products since production costs have dropped significantly. Others are exiting the field entirely because the intellectual stimulation of traditional engineering is gone. If you are feeling the same burnout, it is worth asking whether the boredom comes from the tools themselves or the way the company is forcing
That 90 percent claim dies the moment the repo has secrets. My CI never lets the model see the tests, as the cognitive load has relocated from thinking through system logic to a feedback loop with an LLM.
I'm so jealous of that sabbatical. Taking a full year off to study is the only way to actually avoid burnout. For someone with 16 years in the industry, this transition can feel like a loss of identity. The satisfaction of a perfectly executed architectural pattern is replaced by the relief that the AI did not hallucinate a dependency. If you are feeling the same burnout, it is worth asking whether the boredom comes from the tools themselves or the way the company is forcing you to spend time in a feedback loop with an LLM.
I'm curious if building "cool" hobbies actually scales. Most "cool protocols" just end up as ghost towns within six months. Lately I've been spending more time prompting a model, monitoring its output, and then adjusting the prompt when it falls short—so I wonder whether the lack of sustained interest is a symptom of the tools themselves or the projects we try to launch.
A side app doesn’t beat the noise—even when you start by defining the exact problem in a single sentence upfront, the evening still ends with you untangling the mess. The core issue isn’t just the tool; it’s that the work has become a game of whack-a-mole with an LLM, where the real thinking gets outsourced to a black box. The flow of building something meaningful? Gone. Now it’s just chasing hallucinations and rewriting prompts until something almost works. For someone who’s spent years crafting systems, this feels less like progress and more like watching your craft get outsourced to a glorified autocomplete.
Pure spite. Imagine calling manual coding a time-saver when you can just keep “prompting a model, monitoring its output, and then adjusting the prompt when it falls short,” then blame the AI for the testing failures.