Saturday, August 30, 2025
All the Bits Fit to Print
Reflections on how large language models impact software development and security risks
Martin Fowler shares nuanced reflections on large language models (LLMs) and their evolving role in software development, emphasizing the uncertainty and experimentation still needed to understand their best use. He also discusses the economic bubble surrounding AI and the challenges LLMs bring to software reliability and security.
Why it matters: Understanding how LLMs are actually used in workflows is crucial for accurate assessment of their impact and value in software development.
The big picture: AI-driven software development faces a speculative bubble that could burst, but lasting innovations and value may still emerge.
The stakes: LLMs increase software attack surfaces, especially through agentic browser extensions, posing serious security risks.
Commenters say: Readers appreciate the candid uncertainty about AI’s future, debate the bubble analogy, and highlight the importance of tolerances and repeated questioning when using LLMs.