Web News: Are Web Dev Tutorials Dying?
Two web developers discuss whether traditional web dev tutorials are dying, driven by AI's impact on content creation and learning. They explore how AI is reshaping developer education, the viability of LLM-assisted coding workflows, and the risks of losing human-driven innovation in tutorial content. The conversation raises concerns about junior developers, code quality, and the uncertain future of how programming knowledge is created and consumed.
Summary
The episode is inspired by a video from Kyle of WebDev Simplified, which explored how free tutorial content trains AI systems that then teach users directly, cutting out the original creators and hurting their viewership and revenue. The hosts use this as a launching point to ask whether traditional web dev tutorials are effectively dead and what the implications are for the broader developer ecosystem.
Mike argues that tutorials are dying from multiple angles: prominent creators like Matt Pocock have pivoted from deep technical content (e.g., TypeScript tutorials) to AI workflow content because that is where the audience and revenue now are. He notes this shift is driven not just by AI but also by a declining junior developer job market and broader economic conditions. Companies are pushing employees to upskill in AI, further inflating demand for AI-focused content over traditional syntax-based tutorials.
A major thread of the discussion is the debate over whether programming syntax is 'dead.' Mike draws a distinction between historical abstraction layers in programming (binary → assembly → C → Java → TypeScript), which were deterministic, and AI-generated code, which is non-deterministic. He argues this non-determinism is a fundamental problem because it breaks the predictability that programming depends on, and resolving it by making AI more deterministic would undermine AI's creative value.
The hosts discuss the difficult position of new developers: they must still learn programming syntax to review AI-generated code, but they face social and professional pressure to abandon manual practice in favor of AI-assisted workflows. This creates a paradox where beginners must invest heavily in learning skills they may be told are obsolete, yet without those skills they cannot catch errors in AI output — such as a loop running 30 times unnecessarily.
The conversation also tackles the risk to innovation. The hosts argue that tutorial creators historically drove community-owned innovation — finding creative, unofficial uses of technology that even original developers adopted. They worry that AI trained only on official documentation will lack this creative, jerry-rigged problem-solving capacity, using the analogy of fixing a muffler with a piece of fence rather than the official replacement part.
On the topic of code quality, Mike acknowledges that LLM-assisted development currently produces more inefficient code and introduces more tech debt, but frames it as a conscious bet on the future — accepting current compromises in exchange for workflow fluency that may pay off when LLMs improve significantly. He compares this to the existing risk of human developers pushing bad code, suggesting the difference is one of degree rather than kind.
Finally, the hosts speculate about a future where tutorials are replaced by LLM-ingestible documents, potentially sold through an LLM content marketplace analogous to YouTube, where creators get a cut of tokens spent using their methodologies. They also darkly joke about LLMs eventually creating tutorials for other LLMs, questioning the purpose and direction of the entire ecosystem at that point.
Key Insights
- Mike argues that prominent tutorial creators like Matt Pocock have pivoted entirely to AI workflow content not because of AI alone, but because a declining junior developer job market and poor economic conditions have simultaneously gutted the audience for traditional syntax-based tutorials.
- Mike draws a sharp distinction between AI as an abstraction layer and all prior programming abstraction layers, asserting that the non-determinism of AI-generated code makes it fundamentally different from compilers — and that this breaks the predictability that software engineering depends on.
- The hosts argue that community-driven, unofficial innovation — like jerry-rigging solutions that even original creators later adopt — is something AI trained only on official documentation cannot replicate, representing a concrete and underappreciated loss if tutorial creators disappear.
- Mike frames current LLM-assisted development as a collective 'bet on the future,' where developers and companies consciously accept worse code quality and higher tech debt now in exchange for establishing workflows that may become highly effective once LLM output quality improves significantly.
- The hosts speculate that tutorials may not disappear but transform into LLM-ingestible documents sold through a token-based marketplace, where human creators earn revenue not from video views but from others consuming AI responses built on their documented methodologies.
Topics
Full transcript available for MurmurCast members
Sign Up to Access