I Asked 100 Devs What AI Is ACTUALLY Doing To Their Jobs

Summary

Dr. K reviews developers’ questions about how AI is affecting their work. Many describe losing the enjoyment of programming: instead of solving problems and designing systems, they spend more time prompting AI and reviewing generated code. Others report pressure from managers to produce more in less time, concerns about mediocre software and production incidents, and anxiety about job security.

One developer’s account contrasts years of traditional programming experience with a colleague who used AI heavily and appeared more productive. The example illustrates how AI adoption can affect workplace recognition as well as engineering practice.

Dr. K groups the concerns into three areas:

1. What will happen with AI? Will it replace developers, become more capable and affordable, or prove economically unsustainable? Dr. K says these outcomes are uncertain; neither AI’s future capabilities nor its eventual costs can be predicted confidently.
2. How should developers deal with management? Managers may adopt AI based on hype or expectations of rapid cost savings, while engineers are left responsible for the results.
3. How can developers manage the personal impact? Uncertainty can contribute to burnout, boredom, despair, and loss of motivation. Dr. K argues that maintaining mental resilience and clear judgment is especially important during periods of change.

Engineering Risks and Evidence

• More generated code may mean more defects and instability. Dr. K cites the DORA report, describing a finding that a 25% increase in code output is associated with about a 7% decline in stability. He presents this as a figure to raise in workplace discussions, while acknowledging that he is not an AI technical expert and that much of the available evidence may be around two years old. The estimates may not reflect current models.
• Code review may not fully compensate for AI-generated code. Dr. K discusses research analyzing pull requests and bug-fix outcomes, arguing that visible review activity does not necessarily prevent defects. Large volumes of generated code can also create a review burden: engineers may have to inspect substantial amounts of material without enough documentation or context.
• Human-plus-AI is not automatically better. A meta-analysis of 106 experiments is cited as finding that human–AI combinations performed slightly worse than the better of the human-only or AI-only approaches, though better than a human alone. Dr. K notes that much of the research predates current frontier models, so its relevance is limited.
• AI’s business value is uncertain. Dr. K questions whether short-term productivity gains will translate into profits, especially if AI companies are operating at a loss or if the costs of AI services rise. He presents these as open questions, not settled conclusions.

Practical Guidance for Discussing AI Adoption at Work

Dr. K recommends avoiding a purely oppositional response when management is enthusiastic about AI. Instead:

1. Acknowledge the enthusiasm and match the other person’s emotional tone.
2. Raise concerns constructively, using available data and explaining likely consequences, such as effects on reliability, security, maintenance, and review capacity.
3. Frame risks as planning issues, not as a blanket argument against AI. For example, discuss how increased code output may affect testing, stability, staffing, and costs.
4. Document concerns and expectations, such as in an internal report or email. Dr. K suggests using a SWOT framework—strengths, weaknesses, opportunities, and threats—to present both potential benefits and risks.
5. Avoid “I told you so” responses if problems occur. Seek to recalibrate expectations while keeping records of previously raised risks, and be prepared to respond firmly if blamed unfairly.

The video is primarily a discussion and workplace guide, not a programming tutorial. It also emphasizes that communication and interpersonal skills can help developers influence how AI is adopted, alongside technical expertise.

Main Speakers and Sources

• Dr. K — The main speaker, discussing developers’ concerns and offering workplace and mental-health guidance.
• Developers in submitted questions and anecdotes — Quoted or paraphrased examples from the community.
• Research sources discussed by Dr. K — The DORA report, a study of pull requests and defect outcomes, and a meta-analysis of 106 human–AI experiments.