Martin Fowler on Why he dislikes Large Language Models (LLMs)

Martin Fowler, one of the best authors about software development, recently published an article titled “I Don’t Like LLMs”. The main problem he sees is that they are providing incorrect answers equally confidently as correct ones. Also, their apologies are “fake”.

That’s a valid point, but it doesn’t apply to agents. An agent-based system with a pipeline including a tool for a specific task won’t have hallucinations (which would be obvious), so I wouldn’t make the comparison.