Cheap design
This article explores how advancements in software design, particularly through LLMs, shift the focus from using standard libraries to generating custom code. It highlights the implications for dependency management and emphasizes the need to understand the problem being solved rather than just the mechanics of coding. The author compares this shift to the evolution of 3D printing in manufacturing.
- LLMs flip the standard-library calculus: the default question becomes "is this worth a dependency?" instead of "is there a library for this?", since custom code is now cheap to generate.
- Unlike 3D printing, LLM-generated code has no inherent quality penalty versus hand-written code—if it's correct, it performs identically, unlike physical printed parts that sacrifice strength/precision.
- Design cost (understanding the actual problem/business logic) doesn't disappear even as coding cost collapses—LLMs don't grasp constraints on their own, so that burden stays on developers.
- The tradeoff shifts from dependency/version management to maintenance and vulnerability-tracking responsibility for more bespoke, less shared code.
Design and fabrication costs dictate how physical and software products are built. In the physical world, custom parts are often too expensive, leading to reliance on standard components. 3D printing has reduced fabrication costs but comes with quality limitations, meaning standard parts still dominate in areas requiring precision and strength. In software, design is the main cost since compilation and copying are cheap. This has resulted in a reliance on libraries and frameworks, where developers write glue code to fit these parts together, often leading to bloated and fragile systems.
The emergence of large language models (LLMs) is changing this dynamic. With LLMs, the cost of creating custom code has decreased significantly, allowing developers to generate solutions tailored specifically to their needs without the overhead of managing dependencies. Unlike 3D-printed parts, LLM-generated code carries no inherent quality penalty. If the code is correct, it functions just as well as hand-written code. This shift prompts a new approach to software development: instead of asking if there’s a library available, developers will prioritize whether a problem is complex enough to justify using one.
However, this new model does not eliminate design costs. Understanding the problem remains critical, as LLMs generate code without grasping the underlying business logic or constraints. While the barriers to custom coding have lowered, the responsibility for maintenance and awareness of vulnerabilities now sits with developers, as tracking issues in custom code requires diligence. Ultimately, the new landscape suggests a future where software can be simpler and more efficient, reducing dependency management and accidental complexity.