DiscussionOpinion

Don't Just Vibe Code, Learn While You Code

Matt and Mike discuss how junior developers can learn foundational coding skills while using AI to code, arguing that understanding core concepts remains valuable despite AI automation. They identify three learner camps (vibe coders, aspiring junior devs, and job-focused developers) and debate whether manual coding knowledge is necessary for future success in an AI-driven development landscape.

Summary

The episode explores the tension between using AI coding tools for efficiency and maintaining foundational programming knowledge. Matt initially proposes using AI as a tutor—asking it to explain concepts encountered during development, creating separate educational files, and generating study materials—without reverting to traditional learning methods like MDN or Stack Overflow. He argues this approach improves AI prompts while building understanding.

Mike introduces a critical counterpoint: retention. While following curiosity and learning concepts like for loops is valuable short-term, developers may lose 80% of that knowledge if they don't use it regularly, especially when AI is handling the coding. He emphasizes that job interviews still require manual coding ability, forcing developers to grind both AI prompting skills and traditional coding skills simultaneously—a difficult balance requiring significant time investment.

The hosts identify three distinct learning camps: (1) vibe coders who only want end products and don't care about understanding code; (2) aspiring junior developers improving prompts while learning AI workflows; and (3) goal-oriented learners preparing for jobs requiring both manual and AI coding proficiency. They note that the industry is becoming more competitive, with AI throwing a wrench into an already fast-moving field. Companies conduct interviews with task-based assessments where candidates must demonstrate how they use AI to learn and implement unfamiliar concepts, plus security tests around handling sensitive data.

Both hosts agree that in a utopian world with unlimited time, a 50/50 split between manual coding and AI prompting would be ideal for beginners, but acknowledge real-world constraints make this difficult. They speculate that the next 2-3 years will clarify whether manual coding knowledge remains essential or becomes obsolete as AI systems mature. The discussion concludes with an invitation to listeners to share how they're learning alongside AI coding.

About this episode

AI can write your code, but should it replace learning how to code? Matt and Mike discuss how to use AI coding agents as tutors, learn from the code they generate, and balance traditional programming knowledge with the new skills developers need in an AI-first world.

Key Insights

  • Matt argues that AI can serve as a tutor to explain code concepts in-context, eliminating the need to pause and use traditional learning resources like MDN or Stack Overflow.
  • Mike contends that while developers may forget 80% of learned concepts without regular use, retaining the remaining 20% creates foundation knowledge useful years later and improves AI prompting ability.
  • Matt claims that understanding infrastructure concepts (databases, servers, languages) enables more specific and informed AI prompts, whereas vague prompts from uninformed users yield generic results.
  • Mike observes that most job interviews still require manual coding tests, forcing developers to maintain dual competencies in both AI prompting and traditional code writing.
  • Both hosts recognize that companies now use task-based interviews specifically to assess how candidates use AI to learn unfamiliar technologies, shifting from whether candidates know something to how they ramp up learning.
  • Mike argues that the coding industry gatekeeping is evolving—people with excellent memories or willingness to grind learning schedules will succeed, while others may be excluded due to information retention demands.
  • Matt notes that traditional web development already involved numerous layers of abstraction (webpack, React, Next.js, Svelte), and AI represents another layer that can either teach you or automate away your need to know underlying concepts.
  • Mike states that most current job settings mandate agentic coding and explicitly prohibit manual coding, creating a paradox where developers are forced to use AI while interviews still test manual coding ability.

Topics

AI-assisted learning and prompt engineeringBalancing AI efficiency with foundational coding knowledgeJob interview preparation in the AI eraKnowledge retention and long-term skill developmentThree categories of developers: vibe coders, learners, and job-seekersSecurity and responsible AI use in developmentThe competitive landscape of modern software engineeringInterview assessment methods for AI coding proficiency

Transcript

We've all heard of vibe coding, but what's the next step? I mean, let's just actually back up a little bit. What if you don't want to vibe code at all? What if you're an aspiring junior developer or you're a junior developer looking to level up your skill? And now this AI coding thing has come and kind of jumped on everything. They've it's jostled the entire kind of development world, the software engineering world, and now you're like, well, hang on, do I learn these hard skills? Do I not? Do I spend time learning the hard skills so I can prompt better? Like what, what do I do? But then you have something else kind of nagging…

Full transcript available for MurmurCast members

Sign Up to Access

More from HTML All The Things - Web Development, AI, and Developer Careers

Get AI summaries like this delivered to your inbox daily

Get AI summaries delivered to your inbox

MurmurCast summarizes your YouTube channels, podcasts, and newsletters into one daily email digest.