Code is cheap. Show me the talk.
After 25 years of development, an experienced engineer argues that LLMs haven't just accelerated coding—they've inverted it entirely: code is now abundant and cheap, while the ability to articulate, architect, and evaluate has become exponentially more valuable than syntax knowledge.
Read OriginalMy Notes (1)
"there is an intrinsic value and empathy for the human time and effort that is likely ascribed to it. It is known that there is a physical and cognitive cost that has been paid"
in a lot of places so far there has been a unsaid rule that effort is a proxy for value. one perspective is this adage will completely fall apart due to AI making the need for effort extremely low. there is another perspective which says the due to the need for effort being low, the bar will rise extremely fast and far. the new bar will be quite high, leading to an actual increase in overall effort and taste required to differentiate.
i think the "effort = value" heuristic has always been a shortcut anyway. what people really care about is the outcome, but effort was reliable enough signal that it worked. the second perspective is how i see it playing out. more effort does not mean more "polished" output. my bet is that the "type" of effort that will lead to differentiation will shift upwards. when basic competence becomes trivially cheap, what becomes scarce is taste, judgement, curation and ability to articulate what is actually worth making. the effort, thus moves upstream from execution to discernment
invention of the printing press did not reduce the value of writing per se, however it made "bad" writing more abundant, thus raising the bar for whats worth reading (substack started from a similar thesis)
Summary used for search
TLDR
• Traditional signals of code quality (documentation, commit patterns, comments) are now meaningless—LLMs generate all of them perfectly, making it impossible to distinguish expert work from vibe-coded slop without forensic analysis
• The author personally compresses weeks of work into days/hours while producing better code than manual writing, eliminating the physiological bottleneck of typing that constrained software development for decades
• FOSS collaboration dynamics are fundamentally disrupted: when anyone can instantly generate custom solutions, the incentive to contribute to or use shared libraries evaporates
• The existential crisis for junior developers: those who start with LLM dependency will never develop the foundational skills needed to discern quality or understand systems deeply
• Experienced developers who can "talk well"—articulate problems, architect systems, critically evaluate code—now have disproportionate advantage over those who merely know syntax and frameworks
In Detail
A veteran developer with 25 years of experience argues that software development "as it has been done for decades, is over." The shift isn't about LLMs writing perfect code—it's about eliminating the fundamental constraint that made code valuable: human time and physiological limits. Historically, even with perfect knowledge of what to build, the physical act of typing thousands of lines, maintaining mental maps of complex systems, and coordinating with others made code expensive. That constraint is gone. The author personally experiences 10-100x compression of development time, producing production-grade code better than they could write manually.
This abundance creates a crisis of value and signals. Traditional indicators of quality—thoughtful documentation, consistent commit patterns, well-organized code—are now trivially generated by LLMs in seconds. Without these signals, it's nearly impossible to distinguish expert-crafted systems from vibe-coded slop without deep forensic analysis. The author provocatively suggests that neatly-formatted AI code might actually be more consistent than the "borderline junk" most human developers write daily. What makes code valuable now isn't its technical quality but its provenance: the human accountability, expertise, and articulation behind it. Code that can be infinitely generated without effort is fundamentally hard to value—we're living in a Borgesian Library of Babel.
The implications ripple through the entire software ecosystem. FOSS collaboration, built on the premise that code was expensive to create and valuable to share, faces disruption when experts can instantly generate custom solutions. Junior developers face an existential threat: starting with LLM dependency prevents them from developing the foundational understanding needed to become seniors who can effectively use these tools. The new hierarchy is clear: developers who can articulate problems, architect systems, and critically evaluate code have disproportionate advantage over those who merely know syntax. For the first time, "talk"—in the form of clear thinking and articulation—is exponentially more valuable than the ability to write code.