Engineering management after the cost of code collapsed
karimjedda.com
July 25, 2026
13 min read
🔥🔥🔥🔥🔥
51/100
Summary
The introduction of large language models (LLMs) in organizations has significantly reduced the cost of producing code. Traditional engineering management principles, such as avoiding coding and protecting teams from business pressures, are being reevaluated in light of this change.
Key Takeaways
- The cost of producing code has significantly decreased due to AI advancements, and this change is permanent.
- Many traditional management practices are based on outdated assumptions that need to be re-evaluated in light of the new cost of writing code.
- Metrics like velocity and pull request counts are now misleading because they no longer accurately reflect the effort involved in coding, as the cost of code generation has collapsed.
- Organizations adopting AI must focus on honest instrumentation to assess whether actual productivity has improved, rather than relying on perceived speed.
Community Sentiment
Positives
- If you leverage spec-driven development, LLMs can actually write code better and faster than ever — it's all about guiding them with clear requirements.
- Human peer review is proving invaluable, with real-time discussions enhancing the quality of AI-generated code and ensuring it meets expectations.
- AI has revolutionized autocomplete, dramatically cutting down the time spent searching for documentation and solving common coding problems.
Concerns
- LLMs still struggle with producing maintainable code; what comes out often needs a complete rewrite, raising doubts about their utility in long-term projects.
- The cost of code has actually increased, with code debt piling up faster than teams can handle, contradicting the promise of efficiency.
- Businesses prioritize speed over code quality, leading to a situation where engineers are pressured to push features quickly, often at the expense of solid coding practices.
