Software veteran Martin Fowler sparked a massive debate on Hacker News after admitting he simply doesn't like LLMs. Pointing to their grating, uncanny voice and their knack for delivering confident bullshit alongside genuinely useful answers, Fowler captured a growing developer frustration—while acknowledging we have little choice but to ride the AI train anyway. The surrounding discussion echoed a pragmatic consensus: LLMs are deeply flawed, weird tools, but they offer undeniable leverage.
For early-stage founders, this is a needed reality check. It is easy to get seduced by AI hype or, conversely, to dismiss the tech entirely because of its hallucination problem. But when you are building a company with limited runway, refusing to use a tool that multiplies output just because it occasionally talks nonsense is a luxury you cannot afford. The key isn't trusting LLMs blindly, but leveraging them ruthlessly where execution speed matters.
For small backers evaluating pitch decks, Fowler's critique offers a sharp filter. Be wary of startups whose entire pitch relies on LLMs being magical or flawless. Instead, back the founders who treat AI as a messy, high-leverage workhorse—building smart operational guardrails around flawed technology to reach profitability faster and cheaper than traditional teams.