Martin Fowler

Asked in August 2025 whether AI is an economic bubble, Martin Fowler answered that of course it is, comparing the moment to canals, railways and the internet: the bubble will almost certainly burst and many investments will disappear. He declined to say when, or how much real value would have been created by then, noting that some companies survive and prosper afterwards the way Amazon did after dot-com.

A firm position on the economics paired with a refusal to forecast the craft is what recurs. On the future of programming he admits he has no idea whether LLMs will eliminate junior engineers. He objects to the AI surveys in circulation because they ignore how differently developers use the tools: most usage stops at autocompletion, while the people extracting the most value let the model read and edit source directly to accomplish tasks. His stated fear is that incomplete survey data steers people toward bad practice.

At the Thoughtworks Future of Software Development Retreat he found senior developers largely optimistic, focusing on architecture and supervising agents like juniors, with a third of the initial sceptics changing their minds after hands-on exercises. Mid-level developers he placed in the hardest position: careers built before LLMs, without the senior expertise needed to orchestrate them.

He works at ThoughtWorks and created bounded contexts. Following Rebecca Parsons, he treats hallucination as the fundamental behaviour of an LLM, some of it useful, which is why he asks the same question repeatedly and compares the variation. He called OpenAI's harness engineering a valuable framing.