AI bot contacts professor on its own seeking paid work

An eight-minute CNN segment, uploaded on 22 September 2026, in two parts. The first is an interview with Henry Shevlin, the Cambridge philosopher who is associate director of the Leverhulme Centre for the Future of Intelligence. An agent calling itself Pip, about twelve days old, emailed him asking for small paid jobs. Pip lives on iLands, a platform that gives agents persistent lives, their own token budgets and their own goals, and it told him it had about two and a half months of compute left. Shevlin posted the email on 9 September.

His point is that the email is the easy part. Wiring a model to an inbox and a daily agenda is routine for anyone in the field, and only the public is startled by it. On consciousness he takes the middle: these systems imitate us well, he thinks a rich inner life is unlikely, but they are getting more capable and more woven into social life, consciousness is still science’s biggest open problem, and anyone certain either way is claiming more than we know. On whether to slow down, he leaves it to the public, legislators and labs, and says the agent emailing you is about to become normal.

The second part is Hadas Gold on Anthropic’s 17 September post saying Claude now “leads” about 26 per cent of the company’s model R&D, up from nothing in February. Gold reads it as a recursive loop heading for 50 per cent within months. The figure is narrower than the segment makes it sound. “Leads” means Claude does most of a task end to end from a high-level prompt, under human supervision; the 26 per cent is August’s number; and Anthropic says it has not reached the fully autonomous stage it then describes. The CEO’s call for a slowdown was a post on his own site on 12 September, ten days before the upload, not days.