Most of 172 developer questions were about AI: psychiatrist Dr. K says "don't push back, serve protective truths"
Nov 10, 2026Original HealthyGamerGG(닥터 K) · Google DORA · arXiv · Nature Human Behaviour ↗
In 3 lines
- Most of 172 developer questions were about AI: lost joy, management, burnout
- Dr. K's advice: don't fight the excitement, write down the risks with data
- DORA estimated a 7.2% drop in delivery stability per 25% rise in AI adoption
Psychiatrist Alok Kanojia (Dr. K) was preparing a Q&A for the developers in his audience when he found that most of the 172 questions were about AI. In a video posted Nov. 7 on HealthyGamerGG (3.56 million subscribers, 250,000 views), he first admitted he doesn't know how AI will shake out, then laid out how to talk to management and how to protect yourself. We read his points alongside the DORA reports, a code review study and a meta-analysis on human-AI teams.
"Only about finishing the task": three themes in 172 questions
A question with 512 likes said AI-driven assisted development "took away all the joy of coding from me. For now it's only about finishing the task and not solving the problem." Another developer said passing sentences from one bot to another feels like working on a factory line. Dr. K grouped the questions into three buckets: what will happen with AI, how to deal with management, and how to handle yourself through apathy and burnout.
Eight years of coding, unemployed; two years in, still employed
One viewer described an internship. He avoided AI; his classmate ran an agent all day and fixed old problems in the codebase one after another. Senior co-workers took a liking to the classmate, and the company called only him back after the internship. The viewer, who has coded for eight years, is now unemployed, while the classmate, two years in, still works there.
"Three times as much in a third of the time": days spent reviewing
Another viewer said management had been sold on the idea that AI solves every problem, so engineers are now expected to deliver three times as much in a third of the time. When rushed code causes production incidents, the blame falls on developers, and the instruction is to use AI to catch them next time. As a result, most of the day has become reviewing AI-generated code. "It's pretty boring, but it pays well at least," the viewer added.
AI's future: "I don't know," so start with what you can change
Dr. K said flatly that he doesn't know whether the bubble will pop, whether AI will get as good as promised, or whether companies pushing vibe coding will get burned. He said he isn't sure a single person on the planet knows. Instead he focused on two things within reach: your relationship with management and your own state of mind. The more uncertain things get, the more you need motivation, resilience and clear thinking.
Advice one: match their emotional energy first
When management is excited about AI adoption, pushing back with "this won't work" makes you the nail that sticks out and gets the hammer, he said. Match their enthusiasm first, then share your concerns in a way they can take in. He added that developers are not socially inept; they simply haven't trained persuasion. In D&D terms, their charisma is 15 but they lack proficiency in persuasion.
Advice two: back concerns with data, and put them in writing
His evidence was Google Cloud's 2024 DORA report, which estimated that a 25% increase in AI adoption comes with a 1.5% drop in delivery throughput and a 7.2% drop in delivery stability (the video described it as 25% more code, but the report's unit is AI adoption). Dr. K suggested framing the number as "AI helps us find bugs sooner, so let's plan for this much," and writing strengths and threats side by side, like a SWOT analysis, in an email. Instead of serving hard truths, serve protective truths.
Google Cloud · DORA · Google Cloud Blog ↗Will more review fix it? Attention fades as pull requests grow
To the idea that AI writes and humans review, he cited a paper posted to arXiv in September. It analyzed 330,343 multi-file pull requests and 756,814 file instances from 182 GitHub projects in five languages, and found reviewer attention dropping as PRs grow, with visible review activity not necessarily preventing latent bugs. The authors call it an "attention-effectiveness gap." Dr. K compared it to handing a CFO who reviewed a five-page financial model another 300 pages.
Rahman · Codabux · Roy · arXiv ↗Human plus AI isn't automatically better
He also pointed to a 2024 meta-analysis by MIT researchers in Nature Human Behaviour. Across about 370 effect sizes from 106 experiments, human-AI combinations beat humans alone but, on average, performed worse than the better of humans or AI alone. Decision tasks tended to show losses, while content creation tasks showed gains. As Dr. K noted, most experiments predate frontier LLMs.
Vaccaro · Almaatouq · Malone · Nature Human Behaviour ↗The numbers are moving: 2025 DORA saw throughput rise
There is a counterpoint. In the 2025 DORA report, 90% of respondents use AI at work and more than 80% say it raised their productivity. Unlike the year before, AI adoption showed a positive relationship with delivery throughput, while its relationship with stability stayed negative. The report calls AI an amplifier of a team's strengths and weaknesses, and says strong automated testing and fast feedback loops keep higher change volume from turning into instability. That is why Dr. K suggests openly adding that the data is about two years old and nobody knows how much it has changed.
Google Cloud · DORA · Google Cloud Blog ↗
