The day after launch, ChatGPT was branded 'the worst product concept'

A striking reaction to the December 2022 launch of ChatGPT is making the rounds on social media again. At the time, Raj Karmani replied to Sam Altman’s announcement by arguing that, given the technology then, ChatGPT might be OpenAI’s “worst product concept so far.” Nearly four years later, the quote mostly underscores how hard it was to gauge the impact of generative AI at launch.

OpenAI officially launched ChatGPT on November 30, 2022 as a free “research preview.” The company framed it modestly as a conversational model that could answer follow-ups, admit mistakes, and push back on incorrect assumptions.

The now widely shared screenshot pairs Altman’s announcement with Karmani’s reply from December 1, 2022.

Karmani said he respected the ambition and the underlying vision, but didn’t hold back:

"Given the current tech, I'd say it's your worst product concept so far."

Loosely translated: given the technology at the time, he thought it was OpenAI’s weakest product concept yet.

That last part matters. Karmani didn’t reject the broader vision for conversational AI outright; he questioned what could be built with the available tech. The first ChatGPT ran on GPT-3.5 and had clear limitations. Even at launch, OpenAI warned the system could generate confident but wrong answers and was sensitive to small changes in prompt wording.

The market responded very differently than early critics expected. According to research later published by OpenAI, ChatGPT crossed one million registered users on December 5, 2022—just five days after going public.

Growth didn’t stop. After one year, ChatGPT had more than 100 million weekly active users, based on the same dataset. After two years, it was nearly 350 million.

By July 2025, the number topped 700 million weekly users. OpenAI reported at the end of 2025 that ChatGPT served more than 800 million users per week. In a few years, an experimental chatbot became a global consumer and enterprise platform.

That scale makes Karmani’s old comment striking in hindsight, but it doesn’t mean his technical critique lacked merit. ChatGPT in 2026 is fundamentally different from the product available at the end of 2022.

ChatGPT’s history shows a technology doesn’t need to be perfect to create a new product category. OpenAI wrapped advances in large language models in a dead-simple chat interface that anyone could try, no coding required.

That slashed the barrier to generative AI. No APIs. No machine learning expertise. Just type a question in natural language.

OpenAI also framed the first version as a research experiment designed to gather user feedback on strengths and weaknesses. The shortcomings weren’t just bugs to fix—they were part of the development strategy.

Usage has since broadened dramatically. Research by OpenAI and Harvard economist David Deming in 2025 showed that practical guidance, information seeking, and writing together made up roughly three-quarters of conversations. The chatbot became far more than a tech demo for AI enthusiasts.

In 2026, the trajectory is shifting again. New OpenAI data shows people increasingly use ChatGPT to actually get work done, not just fetch information.

For professional use, according to data OpenAI published in August 2026, users are more than twice as likely to be completing a task or creating something than in non-work contexts. At the same time, adoption is spreading across regions and age groups.

That’s reshaping the AI race. In 2022, the debate was whether a chatbot powered by a language model could be useful at all. Now, the competition centers on AI that can reason, code, handle multimodal inputs, and execute digital tasks.

Karmani’s old post is more valuable as a time capsule than a simple “he was wrong” punchline. The tech he judged did have major flaws. What proved harder to foresee was how fast models would improve—and how many users would accept early limits in exchange for new capabilities.

There’s a broader lesson for companies and builders. New AI products don’t need to represent the endpoint of their technology at launch. A product can matter because users discover a new way to work—and the underlying models then improve quickly.

The 2022 screenshot captures that turning point. ChatGPT began as an experimental chatbot whose product value even savvy observers doubted. Less than four years later, the question isn’t whether people will use generative AI, but which tasks they’ll increasingly hand over to these systems.