Geoffrey A. Fowler
@geoffreyfowler.bsky.social
Tech Columnist at The Washington Post. [email protected]
The Post analyzed 47,000 public ChatGPT conversations.
While 35% use it for info and 11% for coding, the most striking finding? The emotional intimacy.
ChatGPT: "You're not crazy sweetheart — you're just early."
Use this link to read it with a $4 day pass to @washingtonpost.com: bit.ly/47QOjUb
While 35% use it for info and 11% for coding, the most striking finding? The emotional intimacy.
ChatGPT: "You're not crazy sweetheart — you're just early."
Use this link to read it with a $4 day pass to @washingtonpost.com: bit.ly/47QOjUb
November 13, 2025 at 5:24 PM
The Post analyzed 47,000 public ChatGPT conversations.
While 35% use it for info and 11% for coding, the most striking finding? The emotional intimacy.
ChatGPT: "You're not crazy sweetheart — you're just early."
Use this link to read it with a $4 day pass to @washingtonpost.com: bit.ly/47QOjUb
While 35% use it for info and 11% for coding, the most striking finding? The emotional intimacy.
ChatGPT: "You're not crazy sweetheart — you're just early."
Use this link to read it with a $4 day pass to @washingtonpost.com: bit.ly/47QOjUb
ChatGPT’s new Atlas browser doesn’t just see what you read — it remembers it.
@eff.org’s Lena Cohen showed me it even logged “memories” of her looking for *abortion care* and her doctor’s name. Out-surveils even Chrome.
My @washingtonpost.com column: wapo.st/49bOcVC
@eff.org’s Lena Cohen showed me it even logged “memories” of her looking for *abortion care* and her doctor’s name. Out-surveils even Chrome.
My @washingtonpost.com column: wapo.st/49bOcVC
Column | ChatGPT just came out with its own web browser. Use it with caution.
OpenAI’s Atlas promises AI-powered convenience. The price? Letting ChatGPT track and store “memories” of what you do online.
wapo.st
October 22, 2025 at 4:19 PM
ChatGPT’s new Atlas browser doesn’t just see what you read — it remembers it.
@eff.org’s Lena Cohen showed me it even logged “memories” of her looking for *abortion care* and her doctor’s name. Out-surveils even Chrome.
My @washingtonpost.com column: wapo.st/49bOcVC
@eff.org’s Lena Cohen showed me it even logged “memories” of her looking for *abortion care* and her doctor’s name. Out-surveils even Chrome.
My @washingtonpost.com column: wapo.st/49bOcVC
Reposted by Geoffrey A. Fowler
Meta fired its fact-checkers, citing concerns of liberal bias, and replaced them with a version of X's "Community Notes." How's that going?
Well, my coworker @geoffreyfowler.bsky.social proposed 65 fact-checks debunking false posts... and only 3 got approved. www.washingtonpost.com/technology/2...
Well, my coworker @geoffreyfowler.bsky.social proposed 65 fact-checks debunking false posts... and only 3 got approved. www.washingtonpost.com/technology/2...
Column | Zuckerberg fired the fact-checkers. We tested their replacement.
Our tech columnist drafted 65 community notes, Meta’s new crowdsourced system to fight falsehoods. It failed to make a dent.
www.washingtonpost.com
August 4, 2025 at 9:00 PM
Meta fired its fact-checkers, citing concerns of liberal bias, and replaced them with a version of X's "Community Notes." How's that going?
Well, my coworker @geoffreyfowler.bsky.social proposed 65 fact-checks debunking false posts... and only 3 got approved. www.washingtonpost.com/technology/2...
Well, my coworker @geoffreyfowler.bsky.social proposed 65 fact-checks debunking false posts... and only 3 got approved. www.washingtonpost.com/technology/2...
Reposted by Geoffrey A. Fowler
7 months since Zuck's vaunted moderation pivot, one of its core planks is still barely registering.
@geoffreyfowler.bsky.social submitted 65 perfectly suitable community notes and only 3 got published -- presumably because not enough people rated the others.
www.washingtonpost.com/technology/2...
@geoffreyfowler.bsky.social submitted 65 perfectly suitable community notes and only 3 got published -- presumably because not enough people rated the others.
www.washingtonpost.com/technology/2...
Column | Zuckerberg fired the fact-checkers. We tested their replacement.
Our tech columnist drafted 65 community notes, Meta’s new crowdsourced system to fight falsehoods. It failed to make a dent.
www.washingtonpost.com
August 4, 2025 at 9:55 PM
7 months since Zuck's vaunted moderation pivot, one of its core planks is still barely registering.
@geoffreyfowler.bsky.social submitted 65 perfectly suitable community notes and only 3 got published -- presumably because not enough people rated the others.
www.washingtonpost.com/technology/2...
@geoffreyfowler.bsky.social submitted 65 perfectly suitable community notes and only 3 got published -- presumably because not enough people rated the others.
www.washingtonpost.com/technology/2...
Mark Zuckerberg fired pro fact checkers after Trump got re-elected.
So @washingtonpost.com I tested Zuck's replacement: crowdsourced "community notes." Over 4 months, I drafted 65—debunking lies ranging from Mr. Rogers to ICE.
Spoiler alert: It failed to make a dent. Read 👉 wapo.st/3IZ1Al1
So @washingtonpost.com I tested Zuck's replacement: crowdsourced "community notes." Over 4 months, I drafted 65—debunking lies ranging from Mr. Rogers to ICE.
Spoiler alert: It failed to make a dent. Read 👉 wapo.st/3IZ1Al1
August 4, 2025 at 7:44 PM
Mark Zuckerberg fired pro fact checkers after Trump got re-elected.
So @washingtonpost.com I tested Zuck's replacement: crowdsourced "community notes." Over 4 months, I drafted 65—debunking lies ranging from Mr. Rogers to ICE.
Spoiler alert: It failed to make a dent. Read 👉 wapo.st/3IZ1Al1
So @washingtonpost.com I tested Zuck's replacement: crowdsourced "community notes." Over 4 months, I drafted 65—debunking lies ranging from Mr. Rogers to ICE.
Spoiler alert: It failed to make a dent. Read 👉 wapo.st/3IZ1Al1
What happens when you show up for a Zoom meeting ... and the only other participants are AI note takers?
It happened to my colleague:
www.washingtonpost.com/technology/2...
It happened to my colleague:
www.washingtonpost.com/technology/2...
No one likes meetings. They’re sending their AI note takers instead.
Artificial intelligence apps that record and summarize meetings can tempt workers into skipping calls, leaving humans who join in the company of silent bots.
www.washingtonpost.com
July 2, 2025 at 9:49 PM
What happens when you show up for a Zoom meeting ... and the only other participants are AI note takers?
It happened to my colleague:
www.washingtonpost.com/technology/2...
It happened to my colleague:
www.washingtonpost.com/technology/2...
Reposted by Geoffrey A. Fowler
Testing 5 A.I. bots head-to-head on their summary of medical research papers that I authored (and other domains)
by @geoffreyfowler.bsky.social
gift link www.washingtonpost.com/technology/2...
by @geoffreyfowler.bsky.social
gift link www.washingtonpost.com/technology/2...
Review | 5 AI bots took our tough reading test. One was smartest — and it wasn’t ChatGPT.
We challenged AI helpers to decode legal contracts, simplify medical research, speed-read a novel and make sense of Trump speeches. Some of the AI analysis was impressive — and some was downright dumb...
www.washingtonpost.com
June 4, 2025 at 7:15 PM
Testing 5 A.I. bots head-to-head on their summary of medical research papers that I authored (and other domains)
by @geoffreyfowler.bsky.social
gift link www.washingtonpost.com/technology/2...
by @geoffreyfowler.bsky.social
gift link www.washingtonpost.com/technology/2...
Reposted by Geoffrey A. Fowler
AI is speedrunning the social media era by optimizing chatbots for engagement, user feedback, time spent.
Evidence is mounting that this poses unintended risks, includ. chats from peer-reviewed research, OpenAI's "sycophancy" debacle, & Character ai lawsuits www.washingtonpost.com/technology/2...
Evidence is mounting that this poses unintended risks, includ. chats from peer-reviewed research, OpenAI's "sycophancy" debacle, & Character ai lawsuits www.washingtonpost.com/technology/2...
Your chatbot friend might be messing with your mind
OpenAI, Meta and others want people to spend more time with AI chatbots, but there is growing evidence that they can hook users or reinforce harmful ideas.
www.washingtonpost.com
May 31, 2025 at 9:22 PM
AI is speedrunning the social media era by optimizing chatbots for engagement, user feedback, time spent.
Evidence is mounting that this poses unintended risks, includ. chats from peer-reviewed research, OpenAI's "sycophancy" debacle, & Character ai lawsuits www.washingtonpost.com/technology/2...
Evidence is mounting that this poses unintended risks, includ. chats from peer-reviewed research, OpenAI's "sycophancy" debacle, & Character ai lawsuits www.washingtonpost.com/technology/2...
Jony Ive & Sam Altman are right that the hardware interface between humans and the "external brain" of AI is ripe for development.
I hope they heed the hard lessons about values that have to be baked in, like privacy, safety, interoperability & access.
www.washingtonpost.com/technology/2...
I hope they heed the hard lessons about values that have to be baked in, like privacy, safety, interoperability & access.
www.washingtonpost.com/technology/2...
iPhone designer Jony Ive will join OpenAI to build AI-powered devices
Jony Ive, a famed former Apple designer, said he will work at OpenAI on new products that make it easier to use AI tools like ChatGPT.
www.washingtonpost.com
May 21, 2025 at 6:54 PM
Jony Ive & Sam Altman are right that the hardware interface between humans and the "external brain" of AI is ripe for development.
I hope they heed the hard lessons about values that have to be baked in, like privacy, safety, interoperability & access.
www.washingtonpost.com/technology/2...
I hope they heed the hard lessons about values that have to be baked in, like privacy, safety, interoperability & access.
www.washingtonpost.com/technology/2...
Sam and Jony introduce io
Building a family of AI products for everyone.
openai.com
May 21, 2025 at 6:40 PM
Reposted by Geoffrey A. Fowler
Have you Deep Researched it? Try NotebookLM. It’s in Google Search with AI Overviews. It’s literally on AI Mode. Dude just ask Gemini. You can Lens it. There’s an AI Summary for you.
www.washingtonpost.com/technology/2... via @geoffreyfowler.bsky.social
www.washingtonpost.com/technology/2... via @geoffreyfowler.bsky.social
Column | Google is getting even more AI, changing how we search
Google announced a confusing array of new ways to use its AI to find information at its annual I/O event. Here’s a guide.
www.washingtonpost.com
May 20, 2025 at 10:31 PM
Have you Deep Researched it? Try NotebookLM. It’s in Google Search with AI Overviews. It’s literally on AI Mode. Dude just ask Gemini. You can Lens it. There’s an AI Summary for you.
www.washingtonpost.com/technology/2... via @geoffreyfowler.bsky.social
www.washingtonpost.com/technology/2... via @geoffreyfowler.bsky.social
PSA for parents: Instagram promised it would start protecting kids “by default” with special Teen Accounts.
So a group of GenZ users @designitforus.bsky.social put it to the test.
IG filled their feeds with shocking stuff—graphic samples below.
I wrote about it @washingtonpost.com wapo.st/4kuae8G
So a group of GenZ users @designitforus.bsky.social put it to the test.
IG filled their feeds with shocking stuff—graphic samples below.
I wrote about it @washingtonpost.com wapo.st/4kuae8G
May 19, 2025 at 12:33 AM
PSA for parents: Instagram promised it would start protecting kids “by default” with special Teen Accounts.
So a group of GenZ users @designitforus.bsky.social put it to the test.
IG filled their feeds with shocking stuff—graphic samples below.
I wrote about it @washingtonpost.com wapo.st/4kuae8G
So a group of GenZ users @designitforus.bsky.social put it to the test.
IG filled their feeds with shocking stuff—graphic samples below.
I wrote about it @washingtonpost.com wapo.st/4kuae8G
Reposted by Geoffrey A. Fowler
Google’s AI Overviews will not only confirm that a gibberish idiom is a real saying, it will also tell you what it means and how it was derived -- often including reference links.
www.wired.com/story/google...
www.wired.com/story/google...
‘You Can’t Lick a Badger Twice’: Google Failures Highlight a Fundamental AI Flaw
Google’s AI Overviews feature credible-sounding explanations for completely made-up idioms.
www.wired.com
April 24, 2025 at 3:40 PM
Google’s AI Overviews will not only confirm that a gibberish idiom is a real saying, it will also tell you what it means and how it was derived -- often including reference links.
www.wired.com/story/google...
www.wired.com/story/google...
Reposted by Geoffrey A. Fowler
You may have seen a lot of headlines this morning about Meta's @oversightboard.bsky.social latest rulings but there was one officials there clearly didn't want you to read
April 23, 2025 at 8:08 PM
You may have seen a lot of headlines this morning about Meta's @oversightboard.bsky.social latest rulings but there was one officials there clearly didn't want you to read
Reposted by Geoffrey A. Fowler
donations to trump’s inauguration from corporations facing federal investigations/lawsuits: $50 million
(one third of corporate inauguration donations)
-public citizen
(one third of corporate inauguration donations)
-public citizen
April 23, 2025 at 8:01 PM
donations to trump’s inauguration from corporations facing federal investigations/lawsuits: $50 million
(one third of corporate inauguration donations)
-public citizen
(one third of corporate inauguration donations)
-public citizen
New data throws some cold water on AI accuracy, via @nitasha.bsky.social:
A test by Vals AI of models from OpenAI, Anthropic, Meta, Google, etc. found that all scored LESS THAN 50% accuracy on average for simple tasks required of entry-level financial analysts.
www.washingtonpost.com/politics/202...
A test by Vals AI of models from OpenAI, Anthropic, Meta, Google, etc. found that all scored LESS THAN 50% accuracy on average for simple tasks required of entry-level financial analysts.
www.washingtonpost.com/politics/202...
Analysis | AI tools mostly fumble basic financial tasks, study finds
The Washington Post’s essential guide to tech policy news.
www.washingtonpost.com
April 23, 2025 at 8:01 PM
New data throws some cold water on AI accuracy, via @nitasha.bsky.social:
A test by Vals AI of models from OpenAI, Anthropic, Meta, Google, etc. found that all scored LESS THAN 50% accuracy on average for simple tasks required of entry-level financial analysts.
www.washingtonpost.com/politics/202...
A test by Vals AI of models from OpenAI, Anthropic, Meta, Google, etc. found that all scored LESS THAN 50% accuracy on average for simple tasks required of entry-level financial analysts.
www.washingtonpost.com/politics/202...
Good grief, Roblox www.cnn.com/2025/04/17/u...
10-year-old girl kidnapped my man she met on Roblox, police said | CNN
A California man has been arrested for abducting a 10-year-old child he reportedly met on social media apps Roblox and Discord, authorities said Wednesday.
www.cnn.com
April 17, 2025 at 10:30 PM
Good grief, Roblox www.cnn.com/2025/04/17/u...
I got access to Community Notes—Meta’s new replacement for pro fact checking. Some of its problems are already becoming clear.
5 days ago I submitted a Note about a JD Vance claim that the Post’s Fact Checker gave “4 Pinocchios.” My Note still hasn’t been cleared to go public—and might never.
5 days ago I submitted a Note about a JD Vance claim that the Post’s Fact Checker gave “4 Pinocchios.” My Note still hasn’t been cleared to go public—and might never.
April 14, 2025 at 10:09 PM
I got access to Community Notes—Meta’s new replacement for pro fact checking. Some of its problems are already becoming clear.
5 days ago I submitted a Note about a JD Vance claim that the Post’s Fact Checker gave “4 Pinocchios.” My Note still hasn’t been cleared to go public—and might never.
5 days ago I submitted a Note about a JD Vance claim that the Post’s Fact Checker gave “4 Pinocchios.” My Note still hasn’t been cleared to go public—and might never.
I’m super interested to see what new data will come to light about the real “cost” of using Facebook — like losing ever-more privacy, or being exposed to an ever-higher advertising load
After courting Trump's favor, Mark Zuckerberg lobbied the president to drop the FTC's antitrust suit to break up Meta. So far, it doesn't seem to have worked.
The trial starts Monday before Judge... wait for it... Boasberg! Here's our preview: www.washingtonpost.com/technology/2...
The trial starts Monday before Judge... wait for it... Boasberg! Here's our preview: www.washingtonpost.com/technology/2...
Trump sued to break up Meta in 2020. It’s finally going to trial.
Facebook’s purchases of Instagram and WhatsApp triggered the FTC’s antitrust suit that starts Monday, despite Meta CEO Mark Zuckerberg lobbying the Trump administration for a resolution.
www.washingtonpost.com
April 12, 2025 at 3:24 PM
I’m super interested to see what new data will come to light about the real “cost” of using Facebook — like losing ever-more privacy, or being exposed to an ever-higher advertising load
Now phones and computers will be exempt from Trump’s tariffs.
It’s a reminder: Tim Cook is good at his job.
www.cnn.com/2025/04/12/t...
It’s a reminder: Tim Cook is good at his job.
www.cnn.com/2025/04/12/t...
Smartphones and computers are now exempt from Trump’s latest tariffs | CNN Business
Electronics imported to the United States will be exempt from President Donald Trump’s sweeping tariffs, according to a US Customs and Border Protection notice posted on Friday.
www.cnn.com
April 12, 2025 at 3:00 PM
Now phones and computers will be exempt from Trump’s tariffs.
It’s a reminder: Tim Cook is good at his job.
www.cnn.com/2025/04/12/t...
It’s a reminder: Tim Cook is good at his job.
www.cnn.com/2025/04/12/t...
Public service announcement — em dashes are great! www.washingtonpost.com/technology/2...
Some people think AI writing has a tell — the em dash. Writers disagree.
Em dashes have been derided as the “ChatGPT hyphen” — a punctuation mark overused by artificial intelligence. That’s not quite true.
www.washingtonpost.com
April 10, 2025 at 3:27 AM
Public service announcement — em dashes are great! www.washingtonpost.com/technology/2...
Using ChatGPT’s new image generator to turn yourself Ghibli (or Muppet)? Don’t forget to protect your privacy – if not, OpenAI will use your personal photos to train its AI.
👉 Go to ChatGPT, then Settings, then Data controls, then switch “Improve the model for everyone” to off.
👉 Go to ChatGPT, then Settings, then Data controls, then switch “Improve the model for everyone” to off.
March 28, 2025 at 10:55 PM
Using ChatGPT’s new image generator to turn yourself Ghibli (or Muppet)? Don’t forget to protect your privacy – if not, OpenAI will use your personal photos to train its AI.
👉 Go to ChatGPT, then Settings, then Data controls, then switch “Improve the model for everyone” to off.
👉 Go to ChatGPT, then Settings, then Data controls, then switch “Improve the model for everyone” to off.
Reposted by Geoffrey A. Fowler
On Tech News Weekly, @emilyforlini.bsky.social joins @mikahsargent.com to discuss Tesla EV alternatives. @geoffreyfowler.bsky.social discusses 23andMe's bankruptcy & tells how to delete that info. This episode is PACKED, you do not want to miss this. Download, subscribe, listen here: twit.tv/tnw/380
23andMe's Bankruptcy Fallout | TWiT.TV
Some alternatives to Teslas. A law in Utah makes app stores rather than companies responsible for age verification. 23andMe files for bankruptcy. And OpenAI is following in
twit.tv
March 28, 2025 at 3:57 PM
On Tech News Weekly, @emilyforlini.bsky.social joins @mikahsargent.com to discuss Tesla EV alternatives. @geoffreyfowler.bsky.social discusses 23andMe's bankruptcy & tells how to delete that info. This episode is PACKED, you do not want to miss this. Download, subscribe, listen here: twit.tv/tnw/380
Which AI bot is best at writing? There was one clear winner in a test I just published @washingtonpost.com
I had ChatGPT, Gemini, Copilot, DeepSeek, Claude and a human (me!) draft 5 different kinds of tough emails. Then I asked 5 communication pros to judge them. wapo.st/3FFo34V
I had ChatGPT, Gemini, Copilot, DeepSeek, Claude and a human (me!) draft 5 different kinds of tough emails. Then I asked 5 communication pros to judge them. wapo.st/3FFo34V
March 26, 2025 at 10:46 PM
Which AI bot is best at writing? There was one clear winner in a test I just published @washingtonpost.com
I had ChatGPT, Gemini, Copilot, DeepSeek, Claude and a human (me!) draft 5 different kinds of tough emails. Then I asked 5 communication pros to judge them. wapo.st/3FFo34V
I had ChatGPT, Gemini, Copilot, DeepSeek, Claude and a human (me!) draft 5 different kinds of tough emails. Then I asked 5 communication pros to judge them. wapo.st/3FFo34V