Is "vibe coding" going to replace professional software engineering, or are we on the verge of a massive technical debt crisis?

In this podcast clip from The Engineering Room, Dave Farley and Sam Newman dismantle the hype surrounding coding with AI and explain why non-expert software prototyping is creating an unsustainable software development landscape. While tools like Claude Code and various AI coding assistants make it easy to sketch up an idea, the reality of modern software engineering is that AI cannot architect complex systems.

Watch as they discuss why artificial intelligence is actually a magnifier—designed to make experts better rather than replacing them—and explore the rising industry trend of professional developers stepping in as "AI rescue consultants" to fix systems that are collapsing under their own weight.

This clip is taken from this FULL podcast episode: Can AI Replace Software Architects? 

Listen HERE ➡️ https://open.spotify.com/episode/37VpthPlEWiRvD1T7Hzhql?si=e5d9f1a97dd34d64

-

⭐ Only Patreon Supporters get to see the FULL VIDEO Episodes of The Engineering Room, sign up here ➡️ https://www.patreon.com/c/continuousdelivery

-

🗣️ THE ENGINEERING ROOM PODCAST:

Apple - https://apple.co/43s2e0h 
Spotify - https://spoti.fi/3VqZVIV 
Amazon - https://amzn.to/43nkkRl 
Audible - https://bit.ly/TERaudible 

-

🙏The Engineering Room series is SPONSORED BY EQUAL EXPERTS

Equal Experts is a product software development consultancy with a network of over 1,000 experienced technology consultants globally. They increase the pace of innovation by using modern software engineering practices that embrace Continuous Delivery, Security, and Operability from the outset ➡️ https://bit.ly/3ASy8n0

-

Dave on X ➡️ https://x.com/davefarley77
Dave on LinkedIn ➡️ https://www.linkedin.com/in/dave-farley-a67927/
Dave on bluesky ➡️ https://bsky.app/profile/davefarley77.bsky.social

Sam Newman's Website ➡️ https://samnewman.io/ 
Sam Newman on LinkedIn ➡️ https://www.linkedin.com/in/samnewman/

Modern Software Engineering on X ➡️ https://x.com/ModernSoftwareX
Modern Software Engineering on LinkedIn ➡️ https://www.linkedin.com/company/modern-software-engineering
Modern Software Engineering on bluesky ➡️ https://bsky.app/profile/modernswe.bsky.social

-

#softwareengineering #softwarearchitecture #codingwithai #vibecoding

Continuous Delivery

A podcast discussion arguing that vibe coding — using AI tools to rapidly generate software without deep technical knowledge — is fundamentally misleading as a path to production-ready software. While AI coding assistants are genuinely useful for prototyping and making experienced developers more productive, they don't enable non-developers to build scalable, maintainable systems. The conversation highlights how architectural thinking is critical for long-term software health, and that startups built entirely through vibe coding tend to collapse under their own weight rather than achieving mass-market scale. The speakers contend that AI tools amplify expert capability but cannot substitute for foundational engineering judgment.

Why "Vibe Coding" is a Lie (And Startups are Paying the Price)

All we did is accelerate the parts of the process that never needed accelerating in the first place. Are the models great Socratic partners and task runners? Absolutely. And…that’s pretty much where it ends. They didn’t move the needle otherwise. Programming is still just as hard, performance hasn’t improved at all, bugs are even more innocuous and difficult to trace, mental models are deteriorating, juniors are collapsing, seniors are are feeling mental atrophy, and creativity is stagnating. And what do we really have to show for it, truly?
The moment the industry at large accepts this truth…well, you can imagine what comes after that, which is why they’re trying so hard to fine new catch phrases to delay the arrival of that day (cough loop engineering cough).
In my opinion, <a href="https://daily.dev/posts/ai-coding-will-prevent-expertise-the-need-for-ongoing-friction-in-long-term-skill-formation--qiv3i2hlk" target="_blank" rel="noopener nofollow">AI Coding will Prevent Expertise</a> from even forming to begin with.

Vibe code an app quickly to get it to production, then when you hit scaling issues either debug and fix with AI or if it’s too difficult, pay a contracor to fix it.
Who the hell is complaining that they’re struggling to scale their app? That’s the best problem to have and some of the biggest companies (youtube, twitter, instagram, amazon prime video) all have great talks about how it took them lots of highly paid engineers to fix their difficult scaling issues.

“startups built entirely through vibe coding tend to collapse under their own weight” - what do you mean “tend”?
Based on the extremely short history of ~2 years of vibe-coding?
In a matter of 1-3 years there will come an AI so good at coding it could fix all the crappy code previous vide-coding has produced, so I don’t believe pure vibe-coded products are certain to fail. Even <code>Fable</code> is good enough today if proper harness, tailored to the product, is set.

What good is a companion if it does all the work for you, but does it terribly?

In my opinion, AI is best used as a coding assistant that can help you whenever you find yourself struggling to implement a certain feature/function or would like to automate some of the simple introductory-level “busywork” (template setup, et cetera). Even then, best practices involve reviewing all AI generations to ensure the utmost quality, correctness and alignment with objectives.
Fully relying on AI to generate your code, especially entire classes/components, without scrutiny is a step in the wrong direction, as it will eliminate the need for coding-related critical thinking and run the risk of improper, unoptimized code being included in PRs or even future digital product shipments. Human oversight regarding AI generations and incentivizing active human involvement in the coding process with AI-powered assistance whenever most optimal is, as I see it, the right way forward to maintain critical thinking and retain the longstanding “human element” of coding.