That framing landed. People mostly accepted that AI use is uneven, with software work standing out as the place where current models actually earn their keep. A lot of the practical examples split the same way: LLMs are useful for debugging, quick explanations, ad hoc research, rough drafts, and one-off scripts. They get much shakier when asked to replace deterministic systems, own customer support, or make large unattended changes in real codebases. Several engineers described the same pattern in different domains. The model looks competent on mature, well-represented terrain, then falls apart on niche stacks, brittle legacy systems, or anything where hidden edge cases matter.
The sharper point people added is that “AI adoption” now means two different things. One is active use, where someone opens ChatGPT, Claude, Gemini, or an agent because it helps. The other is passive exposure, where AI overviews, support bots, and product features get inserted into software whether users wanted them or not. Commenters were pretty dismissive of treating those as the same thing. If a company swaps a fast deterministic flow for a slower
LLM wrapper, that does not prove demand. It proves management wanted an AI checkbox. That same split showed up in hiring. Candidates now get asked how they use LLMs, but the question often functions less as a skills probe than as a culture tell. The best advice that emerged was to answer concretely, with one example where AI helped and one where it failed, because vague enthusiasm sounds fake and blanket rejection sounds incurious.
The overall read is that the market is in an awkward middle phase. AI is already genuinely useful in some workflows, especially where outputs can be tested and iterated. But broad consumer dependence has not arrived, and a lot of “adoption” is still hype, mandate, or product bundling rather than pull from users. That makes the near-term business question less “is everyone using AI?” and more “which jobs are people voluntarily hiring it for, and which ones are we forcing it into?”