You have 1 article left to read this month before you need to register a free LeadDev.com account.
Estimated reading time: 3 minutes
Top takeaways:
- AI empowers executives: Tools like Claude Code let tech leaders bypass modern syntax barriers and resume coding.
- Prototyping, not replacement: This trend fuels rapid, high-level experimentation rather than threatening engineers’ jobs.
- Structure prevents slop: Spec-Driven Development (SDD) is essential to turn executive “vibe coding” into scalable software.
AI is rejuvenating senior leaders’ interest in coding. The results so far are… mixed.
Krešo Žmak started out as a coder, but as he ascended the career ladder to become Chief Innovation Officer of Croatian tech firm Infobip, his time in front of the terminal was replaced by meetings.
But Žmak has picked up his passion again – thanks to coding tools like Claude Code. He finished his work and told his wife he’d be home in a couple of hours. Three-plus hours later he was still tinkering.
“It was mind-blowing,” he says. “I’m still managing, but not humans anymore. I’m managing these different agents. We are pushing across the board that everybody should start thinking: what will be the next way of managing a combination of agents and people?”
Žmak is far from alone. “As a CS graduate whose first role was as an automation engineer, I never lost the instinct to build,” admits Ori Bendet, vice president of product management at software testing firm Checkmarx. “For years, the gap between my technical foundation and the pace of modern software development made it impractical to act on it.”
Your inbox, upgraded.
Receive weekly engineering insights to level up your leadership approach.
A vibe shift
That has changed in recent months. “Vibe coding, and specifically working with AI agents like Claude, has changed that entirely,” he says. One recent weekend he fixed 11 bugs across a variety of open-source projects and built three apps for his kids.
It’s not just hobby projects either. “A competitive intelligence agent I prototyped as a personal experiment was picked up and productized by my back-office team this week,” he says. “None of that would have happened without AI as a copilot.”
These tools can help rekindle a passion for coding among people who haven’t hit the terminal in earnest for years. However, unlike Jack Dorsey, who slashed 40% of his staff at Block by claiming AI would replace them, the adoption of AI tools by their bosses ought not to worry rank-and-file engineers.
“What I think this unlocks for executives with a technical background isn’t the ability to replace engineers,” says Bendet. “It’s the ability to experiment at a pace and depth that was previously out of reach.”
More like this
Move fast and break things?
The issue is whether those things remain experiments, or if they get pushed into how a business actually works – and what impact they have if they do.
“When executives revisit coding to speed up the development process and start by throwing a couple of prompts at an LLM, it might feel like magic for a few hours, but it won’t take long for the ambiguity inherent to the process to compound,” says Cian Clarke, head of AI at Nearform, an Irish software company.
Clarke worries that can have a deleterious effect on an organization. In my experience, leaders from non-engineering backgrounds often produce Bolt.new or Loveable prototypes and realize quickly that it’s not as simple as clicking ‘deploy now’ to materialize it.”
Bendet sees the bright side though. “We understand the bigger picture: the architecture, the trade-offs, the business problem we’re solving,” he says. “What we’ve historically lacked is the bandwidth to navigate every modern framework, library, and syntax nuance. AI fills exactly that gap.”
Harnessing eagerness
If an executive’s grasp on the architecture of their company’s software stack is perhaps not as strong as they might think it is, there are ways to solve that, says Clarke. “The real value comes when teams capture requirements precisely,” he explains.
Spec-driven development (SDD) becomes key for this new future. “With SDD, detailed functional requirements, constraints and desired behaviour are all defined before a single line of code is written,” says Clarke. “This keeps outputs closer to what’s actually desired, rather than a product of the model’s hallucinations.”
SDD is a useful framework because it doesn’t throw limitations on vibe-coding – which can be useful if done well. Instead it channels that excitement into a scalable workflow. “This means engineers spend less time course-correcting vibe coded slop and more time crafting the desired architecture and user experience,” says Clarke.
London • June 2 & 3, 2026
LDX3 London agenda is live! 🎉
