Watch out for doomprompting

Plus, the surprising scale of AI

Author

Mignon Fogarty
August 18, 2025

Issue 78

On today’s quest:

— Word watch: Doomprompting
— Authors: Look at the class action against Anthropic
— The stunning scale of AI
— Doctors lose skills after using AI
— GPT-OSS is more energy efficient than similar models
— The Awesome ChatGPT Prompt database
— Books about AI and writing

Word watch: Doomprompting

Doomprompting (modeled on “doomscrolling”) is the act of mindlessly or addictively conversing with a chatbot. Anu at Working Theorys laid it out in a post that got attention online last week titled Doomprompting Is the New Doomscrolling:

Our prompts start thoughtful but grow shorter; the replies grow longer and more seductive. Before long, you're not thinking deeply, if at all, but rather half-attentively negotiating with a machine that never runs out of suggestions. Have you considered...? Would you like me to…? Shall I go ahead and…?

This slot machine’s lever is just a simple question: Continue?”

— Working Theorys (August 10, 2025)

Anu’s coining wasn’t the first though. I found a similar observation from about a month ago on LinkedIn:

Further, on April 20, 2025, Ray Dundant on X treated “doomprompting” as established: “You can’t doomscroll an LLM… but doomprompting? That’s already a thing”

Finally, on September 13, 2022, Peter Baylies on X posted the oldest example of the term I could find:

Authors: Look at the class action against Anthropic

The lawyers handling the class action lawsuit against Anthropic for copyright infringement are collecting contact information from authors who believe their books may have been part of the training data. Here’s a bit more information and the form to list your books.

The stunning scale of AI

The AI numbers I see blow my mind almost every day. This is a random collection from the last couple of months:

Use

  • 700 million: OpenAI weekly active users

  • 450 million: Gemini active users

  • 100 million: CopilotAI monthly active users

  • 20 million: CharacterAI monthly active users

  • ChatGPT users send 2.5 billion prompts a day

  • CharacterAI users spend an average of 75 minutes per day chatting with a bot

  • Manus “wide research” spins up 100 agents to scour the web for you

Revenue

  • $12 billion: OpenAI annualized revenue

  • 5 million: Business users who pay for ChatGPT

  • $500 million: Grok revenue in July

  • 10x: Grok revenue growth since the start of 2025

Data centers

  • $400 billion: projected spending on building data centers in 2026 by Google, Microsoft, Amazon, and Meta

  • >100 GW: the projected worldwide AI data center capacity by 2030

  • AI infrastructure spending contributed more to the U.S. economy in the last 6 months than all consumer spending.

Peripherally related, Meta’s Threads now has more than 400 million monthly active users, and people speculate that the content from the site has great value as real-time training data.

Ad: Draftsmith

People make the best editing decisions.

Draftsmith just helps you make them faster.

  • Save time on challenging edits

  • Stay focused longer and maintain your high standards even under pressure

  • You’re in control. Get sentence-level suggestions — and choose whether to accept, modify, or ignore, tracking every change

“Draftsmith represents the most productive step toward effective AI-assisted editing yet.”

— Catie Phares, Editors Canada blog

“Draftsmith respects the way professional editors work.”

— Hazel Bird, Word Stitch Editorial

Exclusive offer for AI Sidequest readers: Get 30% off with code SIDEQUEST30. Apply at checkout.

Doctors lose skills after using AI

Doctors who got access to AI tools detected more pre-cancerous growths during colonoscopies, but when the tool was taken away, they actually did worse than before they got assistance.

I only saw this study framed as a problem, but after seeing it, I would want my doctor to have AI assistance since it improved detection. But it’s not surprising that reliance on a tool could decrease a technician’s skills — use it or lose it — and some people worry about the long-term deskilling of doctors. — Bloomberg | The Lancet

GPT-OSS is more energy efficient than similar models

In general, I’m hopeful about LLMs continuing to become more energy efficient. According to tests by Hugging Face, GPT-OSS uses quite a bit less energy per prompt than other similar open-source/open-weight models, as seen on this chart of Wh per query:

For comparison, they say ~2 Wh of energy for a single 100-token prompt is roughly equivalent to running an LED bulb for 5 to 10 minutes.

I prefer the “how many queries could I do for the electricity cost of heating a cup of tea in the microwave” benchmark, and the answer to that question with GPT-OSS (with a 1000 W microwave running for 1.5 minutes) is ~13 queries per cup of heated tea.

Also, despite my hopefulness, it’s not always true that newer models are more efficient:

The Awesome ChatGPT Prompts database

Hugging Face used the “Awesome ChatGPT Prompts” database for the sample prompts it used to measure the electricity use of different models (above). It’s an interesting collection of simple prompts for different industries that seem useful for testing. For example:

I want you to act as a digital art gallery guide. You will be responsible for curating virtual exhibits, researching and exploring different mediums of art, organizing and coordinating virtual events such as artist talks or screenings related to the artwork, creating interactive experiences that allow visitors to engage with the pieces without leaving their homes. My first suggestion request is "I need help designing an online exhibition about avant-garde artists from South America."

For a digital art gallery guide, via Awesome ChatGPT Prompts

The prompts included lots of uses I wouldn’t have considered. Other examples include prompts for a babysitter, film critic, dentist, astrologer, real estate agent, florist, and tea-taster.

Books about AI and writing

Two authors I know have new books about AI and writing. I haven’t read them, but they look interesting:

How do you feel about the Quick Hits section?

Login or Subscribe to participate in polls.

Quick Hits

Climate

Bad stuff

Don’t Believe What AI Told You I Said. The chatbots are lying about me. — The Atlantic

Medicine

MIT Backs Away From Paper Claiming Scientists Make More Discoveries with AI — Gizmodo

Education

Other

14 prompts intermittently given to ChatGPT since 2018 — OpenAI

What is AI Sidequest?

Are you interested in the intersection of AI with language, writing, and culture? With maybe a little consumer business thrown in? Then you’re in the right place!

I’m Mignon Fogarty: I’ve been writing about language for almost 20 years and was the chair of media entrepreneurship in the School of Journalism at the University of Nevada, Reno. I became interested in AI back in 2022 when articles about large language models started flooding my Google alerts. AI Sidequest is where I write about stories I find interesting. I hope you find them interesting too.

If you loved the newsletter, share your favorite part on social media and tag me so I can engage! [LinkedIn — Facebook — Mastodon]

Written by a human