"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
The Post analyzed a collection of thousands of publicly shared ChatGPT conversations from June 2024 to August 2025. While ChatGPT conversations are private by default, the conversations analyzed were made public by users who created shareable links to their chats that were later preserved in the Internet Archive and downloaded by The Post.
Their conclusions might be due to the fact that people simply don't share their productivity chats.
True, but they also say, "It is possible that some people didn’t know their conversations would become publicly preserved online" so it's likely that many of these chats were not deliberately shared, and therefore should include productivity-related ones as well.
It's unlikely they knew it would become publicly available but they were still deliberately shared.
so it's likely that many of these chats were not deliberately shared, and therefore should included productivity-related ones as well.
That still relies on the assumption that people chose to share their productivity-related chats as often as others.
They were being shared by default at first and had to opt out of sharing when many were archived. Unsure if it was true of these particular ones though.
You always had to click the share button. What they changed is if they show up on Google.
The article doesn't really substantiate thier claim anyways. Data analysis, writing, summarization... even "seeking specific information" all seem like productivity tasks to me.
It seems to use the 10% of chats which are "abstract conversations" as justification.
Anyhow, I wholeheartedly agree that the methodology is CLEARLY flawed: there is no reason to expected shared chats are an appropriate representation of ALL chats.
But, even with the flawed sample, I don't even think it supports the assertion anyways.
Anecdotal, but I haven't found ChatGPT very useful for productivity versus Anthropic or Gemini.
ChatGPT to me feels like a platform, and often yields results commensurate with, being the "household name" AI. Meaning it is most known, but more of a generalist and not great at anything.
That used to be true, but Claude is now getting substantially worse.
I suspect that what's happening is that they are trying to bleed out money in slightly less torrential amounts every month, so they're trying hard to constrain how much resources the thing is allowed to consume, meaning that it started out smart and is now getting steadily dumber over time. I was trying to use Claude for a coding project today, and while I'll admit the questions were complex, it really was remarkably dumb in a way that it didn't used to be.
I've found Sonnet 4.5 requires more context building for bigger code asks to yield the same general quality - but the ceiling seems higher. I can context build with 4.5 and accomplish a larger swath of things without getting stuck spinning on rote garbage whereas with 4 that seemed to happen often.
I've also found that context building using Opus 4.1 (and now 4.5) to be pretty useful, then feed that into Sonnet 4.5 for actual execution gets sort of the best of both worlds while reducing cost.