Writing Code Was Never the Bottleneck (ordep.dev)

Article summary

The article argues that writing code is not the bottleneck in software engineering, but rather the processes surrounding it, such as code reviews, knowledge transfer, and testing. The author claims that the use of Large Language Models (LLMs) to generate code does not remove these bottlenecks, but rather shifts the workload. The article suggests that understanding code is still the hard part, and that LLMs do not fix the fundamentals of software engineering. The author concludes that the cost of making sense of code together as a team is still the bottleneck.

Main themes

• software engineering bottlenecks
• LLMs and code generation
• code understanding and maintenance
• industry trends and funding
• productivity and tooling
• market demand and saturation

What commenters say

• Some commenters agree that writing code was never the bottleneck, and that other factors such as market demand and funding are more significant.
• Others argue that LLMs are removing the bottleneck of writing code, but only for non-coders, and that competent coders with LLMs still produce vastly better outcomes.
• A few commenters believe that the industry has become saturated with capital, but lacks good ideas to fund, while others disagree, citing unmet demand for good software in developing countries.
• Some argue that investors prioritize high returns on investment over the actual utility of what gets built, and that this can lead to a lack of funding for sustainable businesses that serve niche markets.
• There is disagreement over whether LLMs are increasing productivity, with some attributing improvements to better tooling and others to the models themselves.
• A few commenters note that the capability to write high-quality code is still a scarce resource, and that the industry has begun to care less about this.
• Some argue that the economy is bottlenecked in complex ways, and that it is not obvious that 100x more code would be useful or produce 100x as much value.