From viral videos to offers from companies
What had begun almost by chance, as a pastime for a creator busy working as a gardener, has turned into a showcase of chatbots' limitations. Speaking on Wednesday at the WIRED World Fair, according to WIRED, Matt Husk said that "all the big AI companies" had contacted him to collaborate after the success of his videos, in which he gives the models apparently simple requests — counting to 100, beatboxing — and shows their errors. Among the requests he received, Husk also cited one from Elon Musk's xAI, which he said had offered him a Tesla car for a week in exchange for videos made exclusively for the company.
It should be noted that these are statements by the person concerned reported by WIRED, not facts confirmed by the companies: OpenAI, Anthropic and xAI did not respond to requests for comment. In the available information, the same companies appear under changed names in the source text — xAI is said to have been merged into SpaceX and then renamed following an alleged intervention by President Trump on the definition of artificial intelligence — details that the material reports without further verification.
The timer that gets it wrong and the contacts with OpenAI
Husk's success rests on a simple type of content: putting ChatGPT's voice to the test with measurable tasks and showing, without commentary, the off-target responses. One episode that became emblematic concerns the timing of a run: when asked to time a mile, the model reportedly described an interval of a few seconds as "about 10 minutes." Also according to WIRED, in April OpenAI CEO Sam Altman, a guest on the Mostly Human podcast, reportedly admitted with a smile that the model's inability to set an accurate timer is a "known problem" that is set to last another year.
On the personal side, Husk also said at the WIRED World Fair, interviewed by Evan Ratliff, that OpenAI had offered him early access to a fix for the timer problem in exchange for his account credentials, a proposal he rejected. The company had reportedly previously asked him to collaborate on a video, with a financial offer that Husk described as minimal, and then reportedly offered to send him gifts while asking for his home address, also to no avail in that case.
An economy of verification and the limits of the case
The story also has a professional angle: according to WIRED, Husk has about two million followers in total across Instagram, TikTok and YouTube, an agreement with the Unicorn agency and commercial collaborations with CeraVe and Ninja, as well as an upcoming original series with The Onion in which he will learn skills such as mime via Zoom. He explained that he wants to broaden his scope because he spends too much time talking with AI and that he still prefers his own ideas, even the worst ones, to those generated by the models.
Several elements remain to be verified: the size and nature of the companies' offers, the actual existence of the described communications, and the scope of denials or confirmations, which have not arrived so far. The material also does not make it possible to establish how far the Husk case reflects a structural phenomenon in relations between AI companies and critical creators, nor to quantify the impact of his videos on public perception of the models. The news should therefore be read as a report of public statements at a WIRED event, with the caution that the absence of company responses imposes.
. The statements are attributed to the source; any information limits: All information comes from WIRED and from Matt Husk's statements at the WIRED World Fair; OpenAI, Anthropic and xAI did not comment. The material contains references to an alleged corporate restructuring of xAI and to a redefinition of AI by President Trump reported without independent verification. No data are available on the size of the offers, the actual number of contacts or the measurable impact of the videos..
