Matt Husk says AI labs sought collaborations after his chatbot takedowns

Matt Husk built an audience by asking chatbots to do ordinary things and recording their failures. Now, he says, their makers want to work with him. Speaking with Evan Ratliff at WIRED World Fair on November 4, the creator named OpenAI, Anthropic and xAI as companies that had approached him about collaborations.

Simple requests, revealing failures

Husk’s videos put ChatGPT’s voice model through requests such as counting to 100, beatboxing and timing a run. His deadpan exchanges have helped him accumulate about 2 million followers across Instagram, TikTok and YouTube, according to WIRED’s November 5 account of the interview.

The timer video illustrates the appeal. Husk asks ChatGPT to time him running a mile. After only a few seconds, it tells him that around 10 minutes have passed. When he protests that he was faster, the chatbot responds sympathetically: “It can feel faster in the moment.” The exchange pairs an incorrect measurement with a reassuring explanation.

Shown that clip on the Mostly Human podcast in April, OpenAI CEO Sam Altman acknowledged the timer problem as a “known issue” and said it would remain one for another year.

“The way they promote it is like ‘It can do anything,’” Husk said, “so I’ll put it to the test—see if it can order bread for me.”

Matt Husk, speaking with Evan Ratliff at WIRED World Fair

A timer patch, gifts and a week with a Tesla

Husk said OpenAI offered him early access to the timer patch if he shared his account. He declined. His account describes an offer of access to a fix, not an announcement that the timer problem had been resolved for everyone using ChatGPT.

He also described an earlier email from OpenAI asking him to collaborate on a video. “They offered me the smallest amount,” he said. “I was just like, ‘No.’” He gave no dollar figure in the account. The company then offered gifts and asked for his home address; he said he turned that down, too.

The xAI approach carried a different condition. Husk said the company offered him use of a Tesla for a week in exchange for making videos exclusively for it. He named Anthropic among the companies seeking to work with him, but did not describe a comparable offer from that company in WIRED’s account.

OpenAI, Anthropic and xAI did not respond to WIRED’s requests for comment.

An audience he did not expect

Husk told Ratliff that he began publicly trolling chatbots about a year ago. He had been making TikToks asking people questions, but encountered few people while working his landscaping job and ran out of ideas. ChatGPT supplied someone—or something—to question when other people were not around.

In one early recording, he told ChatGPT a nuclear bomb had gone off. It responded with encouragement and advice to take shelter, which he found unhelpful. Husk thought the video was poor, but said it had reached 100,000 views by the next morning. His response was to try another scenario: telling the chatbot he had been hit by a train.