When ChatGPT took off a few months ago, it began a great wave of experimentation accompanied by wonder and fear at all of its impressive capabilities. In the intervening period, a growing chorus of disappointment and frustration has arisen — along with a widely held sense among avid early users that the bot is getting “dumber.” It can do less, and its answers are more scripted, less creative and revelatory, the thinking goes. Whether the product has actually been throttled back in some way — “nerfed,” as gamers would have it — is a subject of intense ongoing debate. (OpenAI, the company that built it, insists ChatGPT has not been nerfed.) Tech columnist John Herrman sees the question as interesting not just on its own terms but as a lens onto something bigger happening with the world’s most famous AI chatbot: It is shedding its pretensions of being a stand-alone technological marvel and taking its place as a killer — and very lucrative — feature of Microsoft’s productivity-software suite (you know, Word, Outlook, PowerPoint, and so on). Google’s AI chatbot appears to be on a similar course. So maybe those wide-eyed earlier adopters, wowed and full of wonder (or fear), were just cogs in the final stages of a complex software-development plan? As John puts it, “They were helping with market research and marketing, and their work is nearly done.”