Developers describe AI coding’s tradeoff: faster output, less hands-on satisfaction
CNBC’s new interviews expose a tension: engineers still need judgment and experience, even as AI takes over work that helped build both.
Loading page…
CNBC’s new interviews expose a tension: engineers still need judgment and experience, even as AI takes over work that helped build both.
Listen to this story
In interviews published November 20, developers described AI coding shifting effort from writing software to reviewing machine-generated work, with uneven effects on job satisfaction and training. Taktile CTO James Broadhead said he can produce 10 prototypes in an afternoon instead of spending three weeks on one, but called that his experience, not an industry-wide productivity measure. Some developers welcome more time for difficult problems; others miss the satisfaction of solving them themselves. The practical challenge is keeping engineers skilled enough to judge AI output and troubleshoot without it.
Broadhead said agents generate enough code that engineers may not read it all, making it harder to maintain a mental model of the software.
Gemma Rodgers described AI-assisted wins as “a breadcrumb of joy, not a full meal of joy,” while also saying the tools can open coding to more people.
Chris Sees treats AI like a capable junior engineer whose work needs supervision and worries automation could remove routine tasks where junior developers gain experience.
Engineers who once spent three weeks building a prototype can now generate 10 in an afternoon, Taktile CTO James Broadhead told CNBC. But choosing what deserves to survive is harder. In interviews published November 20, developers described a job shifting toward judging machine-written code, with consequences for satisfaction and how newcomers learn.
The accounts capture different experiences of the same tools, not a unanimous verdict against them. Some interviewees welcome having less code to write themselves. Others miss the challenge. Even supporters worry about preserving the experience needed to supervise AI’s work.
For Broadhead, the chief technology officer of AI decision-making company Taktile, the difficult part is deciding which quickly generated options are any good. His prototype comparison describes his experience, rather than an industry-wide productivity measurement.
He also described a loss of familiarity with the underlying software. AI agents generate so much code that engineers no longer read all of it, he said. Previously, building software helped developers form a mental picture of its structure. That becomes harder when the code changes constantly.
Joshua Wöhle, CEO of AI training company Mindstone, sees a widening gap between workers who use AI effectively and those who do not. He believes employers will increasingly face a choice between helping people catch up and prioritizing those already getting the most from the tools.
His explanation also acknowledges the personal cost: people who excelled because they loved working through code are not necessarily today’s strongest performers. The activity they most enjoyed may no longer be what delivers the most value to their employer, in his view.
For U.K.-based developer Gemma Rodgers, the reward used to come from piecing together a solution, sometimes after days or weeks chasing a stubborn bug. AI has reduced some of that satisfaction. Yet she also credits the tools with opening the field to people who might previously have struggled to thrive.
Chris Sees, a software engineer with more than 20 years’ experience and CEO of virtual office company Hoxton Mix, finds satisfaction elsewhere. He told CNBC that AI frees time for interesting problems. He treats it as a capable junior engineer whose work still needs supervision, and enjoys spotting where it has gone wrong.
Sees worries that automating routine coding and debugging could remove learning opportunities for junior engineers. Those tasks, he said, are where developers build experience. Companies need to make AI part of learning rather than a shortcut around it.
Rodgers raised a related practical concern: engineers must retain the ability to investigate problems themselves when AI tools or internet services go down. She worries that pressure to keep producing work is eroding skills developers previously learned on the job.
For her own work, self-employment leaves Rodgers room to choose. She uses AI for tasks she finds tedious while keeping more interesting problem-solving for herself. That control lets her decide which difficulties to automate and which are worth working through.
Loading discussion...
Join the conversation
Explain when the learning is worth the extra time.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.