OpenAI ChatGPT-5 introduction displayed on smartphone screen

(Cheng Xin/Getty Images)

Here’s how people are actually using ChatGPT

The new report released in collaboration with the National Bureau of Economic Research offers one of the largest surveys of real-world AI chatbot use.

Jon Keegan

9/15/25 2:51PM

OpenAI has released its largest report yet on how real people are actually using ChatGPT. The fascinating working research paper, published by the National Bureau of Economic Research, describes a wide-ranging study that used AI to analyze 1 million chat transcripts (no humans read any of the chats). The study has not been peer reviewed.

Some of the big takeaways from the paper:

💃 70% (!!!) of all queries were not related to work. That number may send a chill down the spine of Big Tech, as it’s betting on enterprise AI to generate enough revenue to justify the hundreds of billions it’s spending to build out AI infrastructure.
📝 Among work-related messages, the most common use for ChatGPT is writing, and mostly just to modify or improve a user’s text. Writing queries made up 42% of work-related messages and 52% of all messages from users who work in business and management.
🙋🏻 About half (49%) of all queries were classified as “asking” — for guidance, advice, or information. 40% of messages were requests classified as “doing,” or asking the chatbot to complete a task.
👩‍💻 Female users contributed more than half of all queries, as of July 2025. This is a massive shift from early on, when the vast majority of users were male. But it’s worth noting that the study determined this by classifying first names as masculine or feminine.
🛹 The youth loves AI. Half of all messages were from adults under 26.

The OpenAI researchers took a random sample of about 1 million messages between May 2024 and June 2025 from logged-in, adult ChatGPT users (who did not opt out of sharing their messages for training).

ChatGPT usage - Breakdown of tasks by topic. — Breakdown of tasks by topic (Chart: OpenAI/NBER)

This study is one of the largest surveys of real-world AI use, so this data will be of great interest to all the companies trying to figure out how they’re going to make money selling AI services.

One thing that stood out was how utilitarian the usage of AI was. Rather than falling in love with an AI chatbot or having deep conversations with your new AI buddy, it looks like people are just using it to make their work better and figure things out.

It remains to be seen how AI will end up being part of our everyday lives, but it might look a lot more boring than Silicon Valley is making it out to be.

Rani Molla17h

Tesla and SpaceX to jointly run “most epic chip-building exercise in history by far”

In the latest instance that Elon Musk views Tesla and SpaceX as effectively one company, the CEO of both announced Saturday that the two firms will join forces on his Terafab project — what Musk says will be “the most epic chip-building exercise in history by far.”

Many of the details mirror what we reported last week, with one major addition: SpaceX will play a leading role.

Terafab, whose location is still under consideration as it the facility would be too big to fit on the Giga Texas campus, aims to vertically integrate the entire chipmaking process, from design and fabrication to testing and packaging. The goal is to supply AI chips to Tesla, SpaceX, and its subsidiary, Musk’s AI company, xAI, whose suppliers Musk said will be unable to handle their demand in “three or four years.” While Tesla has designed its own chips, it has never manufactured them.

Musk said the facility is intended to produce up to 1 terawatt of compute annually. The plant will manufacture two types of chips: inference chips for Tesla’s robotaxis and Optimus robots, and custom AI chips intended for space-based applications like solar-powered AI satellites. According to Musk, roughly 80% of the compute will be allocated to space-related uses, with the remaining 20% supporting projects on Earth.

Morgan Stanley has estimated the project could cost Tesla an additional $35 billion to $45 billion in capital expenditures, though now perhaps some of that capex might be shared with SpaceX. Like many of Musk’s ambitions, the project is enormous in scale and will likely to take years to complete — potentially into the end of the decade or beyond.

TERAFAB: the next step to becoming a galactic civilization

Together with @SpaceX & @xAI, we're building the largest chip manufacturing facility ever (1TW/year) – combining logic, memory & advanced packaging under one roof.

To harness as much power as possible from the Sun, we… https://t.co/QYfGR8XsJx
— Tesla (@Tesla) March 22, 2026

Jon Keegan3/20/26

White House releases AI legislative framework

The White House has released its policy wish list for AI legislation — and what it wants excluded.

Still, the odds of any actual AI regulation getting passed in Congress right now are very slim.

The “National Policy Framework” for AI lays out seven issues that the Trump administration wants to see reflected in any congressional action around AI.

The items listed in the framework include:

Child safety protections, age verification, and parental controls for AI.
Data center projects voluntarily pay their own way when it comes to power, but incentives should still be encouraged.
Copyright laws should allow for training models on copyrighted works, while protecting individuals’ voice and likeness.
Free speech should be defended for AI systems, preventing the government from pressuring companies to ban or alter content based on partisan agendas.
A light touch to regulation to encourage innovation, and no federal agency to regulate AI.
American workers vulnerable to AI job replacement should be retrained and supported.
Federal AI rules should preempt any state AI legislation to prevent a patchwork of laws that companies would hate.

The policy list is the latest in a series of proposals from the AI-friendly Trump administration.

President Donald J. Trump Unveils National AI Legislative Framework

The items listed in the framework include:

Child safety protections, age verification, and parental controls for AI.
Data center projects voluntarily pay their own way when it comes to power, but incentives should still be encouraged.
Copyright laws should allow for training models on copyrighted works, while protecting individuals’ voice and likeness.
Free speech should be defended for AI systems, preventing the government from pressuring companies to ban or alter content based on partisan agendas.
A light touch to regulation to encourage innovation, and no federal agency to regulate AI.
American workers vulnerable to AI job replacement should be retrained and supported.
Federal AI rules should preempt any state AI legislation to prevent a patchwork of laws that companies would hate.

The policy list is the latest in a series of proposals from the AI-friendly Trump administration.

SPACE

CASE

Rani Molla, Nate Becker3/20/26

Jon Keegan3/20/26

WSJ: OpenAI rolling everything into one desktop “superapp”

OpenAI is trying to eliminate distractions and focus on building AI that helps with enterprise productivity tasks like coding and organizing spreadsheets.

As part of that effort, the startup is consolidating some of its “side quests” into one “superapp,” according to a report from The Wall Street Journal.

The plan is to merge ChatGPT, Codex, and the Atlas browser together, as it seeks to focus its efforts as it competes with Anthropic and Google for lucrative enterprise customers.

OpenAI Head of Apps Fidji Simo told staffers in an internal memo that “we realized we were spreading our efforts across too many apps and stacks, and that we need to simplify our efforts. That fragmentation has been slowing us down and making it harder to hit the quality bar we want,” per the report.

Exclusive | OpenAI Plans Launch of Desktop ‘Superapp’ to Refocus, Simplify User Experience

The plan is to merge ChatGPT, Codex, and the Atlas browser together, as it seeks to focus its efforts as it competes with Anthropic and Google for lucrative enterprise customers.

Rani Molla

Semi charmed

3/20/26

What we know about Tesla’s Semi truck as production nears

Musk says “high-volume production” will start this year. Wall Street thinks that’s more like 5,000 to 15,000 trucks.