Credit: jackpress via Shutterstock.com
“They are racing straight to self-improving superintelligence and gambling with our lives,” Jacob Coxon wrote this week as he resigned from Anthropic.
He meant it as a serious warning. Within a day, the line had become a joke template on X. Dozens of users copied the structure of his post and swapped the threat of human extinction for something absurd.
The resignation itself was not a joke. Coxon, a former Anthropic researcher who previously worked at OpenAI, said neither Anthropic nor OpenAI was acting responsibly.
He said the people building AI privately feared it could cause human extinction by the end of the decade. He wrote that AI would “soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources”. His post passed 150 million views.
Coxon told Axios he had left Anthropic after about four months, two months before his equity would have vested. Then the imitations began, Business Insider reported.
The format
The joke follows a fixed shape. The writer resigns dramatically.
They accuse their employer of racing recklessly toward disaster. Then they reveal that the supposed catastrophe is something harmless or silly. The reveal is usually a single swapped phrase, which is what made the format so easy to copy. The grave tone of Coxon’s original stays intact. Only the danger changes.
Many of the posts came from people who work in AI and technology themselves, rather than from outside critics. Some kept Coxon’s opening words verbatim before diverging; others rebuilt the whole passage around a new punchline.
The contributors ranged from designers and software engineers to startup founders.
Jae Holzman, a senior reporter at Heatmap, wrote that they had resigned from the Acme Corporation after years of “pre-gag research” at Acme Paint. In a close riff on Coxon’s wording, Holzman said Acme was racing toward “self-painting fake tunnel technology”.
The reference is to the Looney Tunes cartoons. In them, Wile E. Coyote paints a fake tunnel on a rock face.
The Road Runner runs through it, and Coyote crashes into the wall.
Pop lyrics and spreadsheets
Other posts borrowed from music. The designer Isha Taghvaei published a widely viewed post that opened with Coxon’s exact first line, then switched to lyrics from Natasha Bedingfield’s 2004 song “Unwritten”.
In place of self-improving superintelligence, Taghvaei wrote that OpenAI and Anthropic were “releasing their inhibitions” and “feeling the rain on their skin”. The software engineer Alan Rice used the same template, then pivoted to “plants and birds and rocks and things”.
That line is from “A Horse With No Name”, the 1971 song by the band America.
Startup workers found their own fictional dangers. Ian Tracey, a senior staff software engineer working on applied AI at the fintech company Ramp, joked that his employer was racing toward “self-closing books”. He warned there was no plan for what happens when businesses save too much time and money.
Shane Hegde, the chief executive of Air, warned of a self-organising workspace that might let creative teams be creative. Kabir Sethi, a growth lead at Phoenix Trade, adapted the format for crypto. He wrote that his employer was racing toward “seamless global trading of any asset” and gambling with his money.
How far it travelled
The original post reached a large audience before the parodies took hold. It passed 150 million views, according to Bloomberg, and reached more than 100 million people within days, according to Wired.
The scale of the response drew comment from officials. Emil Michael, the US Defense Department’s chief technology officer, dismissed the warning at a defence conference, describing it as “all the fears that have coalesced in one well-written tweet”, Bloomberg reported.
Most of the imitators, by contrast, worked in technology and treated the format as an in-joke about their own industry.
The fear underneath the joke
The template is a joke, but the warning it copies is not. Coxon said leading labs privately fear that advanced AI could cause human extinction within a few years. His resignation drew public support from senior researchers.
Evan Hubinger, who leads alignment stress testing at Anthropic, said he believed there was a greater than 10 percent chance AI could kill all humans within the next decade.
Anthropic researcher Samuel Marks wrote that AI developers believe the technology could cause human extinction.
Others outside Anthropic have offered similar figures. Geoffrey Hinton, one of the researchers often called a godfather of AI, has said a 10 percent chance of AI killing humanity is not unreasonable. Toby Ord, the Oxford researcher and author of “The Precipice”, puts the risk at about one in 10 by 2100.
Not everyone agrees.
Gary Marcus, an AI researcher, has said he knows of no realistic scenario for actually killing all humans. Roman Yampolskiy, the author of “Artificial Superintelligence”, has put the chance far higher, above 90 percent if a general superintelligence is built. Those competing estimates are the reason the original post spread, and the reason the parodies of it landed.
The debate behind it
The wider debate has been running for months.
In July, about 1,400 employees across frontier AI companies signed an open letter. It urged the US government to build tools to slow the pace of AI development. In the same period, the United Nations human rights chief called for red lines on the technology. OpenAI’s chief scientist Jakub Pachocki wrote that the moment called for extreme caution.
The AI extinction meme took that anxiety and, for a day, made it funny.
Get the TNW newsletter
Get the most important tech news in your inbox each week.
