ChatGPT launched free, then Plus arrived ten weeks later

The writer's problem

Any account of how AI writing tools spread across drafting, editing and publishing over the past several years tends to point back to a single release, and getting that reference right means stating precisely what was actually announced on that date, rather than crediting it retroactively with capabilities the product only gained months or years later.

What the documents show

OpenAI's own post, Introducing ChatGPT, dated 30 November 2022, describes a model trained to interact in a conversational way, able to answer follow-up questions, admit mistakes, challenge incorrect premises and reject inappropriate requests, released as a free research preview so the company could, in its own words, get users' feedback and learn about its strengths and weaknesses. The same post lists limitations directly, stating the model sometimes writes plausible-sounding but incorrect or nonsensical answers, and notes it was fine-tuned from a GPT-3.5-series model that finished training in early 2022. A second OpenAI post, Introducing ChatGPT Plus, dated 1 February 2023, announced a 20-dollar-a-month subscription and stated the company would continue to offer free access to ChatGPT, explicitly describing the original release as a research preview and citing professional use cases including drafting and editing content among the feedback received in the interim. Read together, the two OpenAI documents place the free public research-preview release on 30 November 2022 and the first paid tier on a separate date roughly nine weeks later.

The editorial choice

An entry describing the ChatGPT launch as an origin point for later writing-tool adoption should specify which document it actually means. The free research-preview post is the correct primary source for the initial public release; monetization, later model versions, and features added afterward belong to their own separately dated announcements, not to the 30 November 2022 post. This is a discipline point worth stating as editorial practice: dating a claim to what a company actually announced on a given day, rather than to what later commentary attributed to that day, is what keeps a retrospective entry accurate rather than merely convenient to write.

What stays with the author

OpenAI's own November 2022 post names its model's central limitation in terms directly relevant to anyone using it for writing: producing plausible-sounding but incorrect text, a problem the company states is difficult to fully resolve through its own training process. Whatever value a writer draws from a general-purpose chat assistant for drafting or brainstorming, verifying anything resembling a fact, a quote or a citation before it reaches finished writing sits outside what either of these two OpenAI documents claims the tool itself solves.

• Which specific document is actually being cited when a piece refers to the ChatGPT launch?
• Does a claimed capability belong to the original research preview or to a later, separately dated update?
• Has anything resembling a fact, quote or citation in the drafted text been checked against an independent source?

Two OpenAI posts, nine weeks apart, mark a free research preview and its first paid tier as distinct events, and collapsing them into one undated launch loses exactly the distinction a careful retrospective needs to keep.