Surgeons at the National Hospital for Neurology and Neurosurgery in London removed an 11mm tumour from a patient's pituitary gland. At the same time, an AI system analysed the live endoscope feed, identifying nerves and blood vessels to avoid.

The operation took place in May and was announced on Thursday, once Rhys Hibbert, 48, had recovered. It saved his sight, and he was walking unaided within a week.

Two details matter more than the milestone itself. The surgical team stayed in full control throughout, so this is real-time segmentation on a camera feed rather than a robot operating. And the system was built in-house at the UCL Hawkes Institute, running on Nvidia's Clara IGX platform, funded by the National Institute for Health and Care Research alongside Google. It didn't appear out of nowhere.

The same group, with Hani Marcus as senior author and Danyal Khan leading, published on AI-assisted anatomy recognition in endoscopic pituitary surgery in npj Digital Medicine back in 2024. This is a research pipeline reaching a patient rather than a demo.

One caveat worth noting: the registered trial protocol circulating alongside the story describes AI output being displayed on tablets positioned for surgical residents and nurses rather than the lead surgeon, with a stated educational purpose.

The Guardian reported the operation and Hibbert's account of waking able to see the room clearly. Digital Health set out the technical stack and the NIHR and Google funding behind the trial. An analysis at explainx.ai stresses this was segmentation on an endoscope feed rather than an autonomous surgeon, and is careful to say it cannot confirm which registered trial the operation falls under.

Notice what actually shipped. Not an autonomous surgeon, but a model that watches a camera and labels what a human should not cut, with the human keeping the instrument. That is the shape most useful AI takes in regulated, high-consequence work, and it is the same pattern running through everything we have covered this month, from coding agents taking on legacy migration to digital agents handling routing before a person picks up.

The line worth South African attention is the one about countries with limited surgical expertise. We have very few neurosurgeons for the population, and they sit in a handful of centres, so software that lowers the expertise threshold on the hardest procedures is worth more here than in London.

The open question is which regulator signs off a model like this, and on what evidence. Better answered before the product exists than after.

You might also like our piece on why SA enterprise AI adoption keeps outrunning any strategy for it, Anvaya's seed funding for AI inside a regulated profession, and how developer career frameworks change when AI takes the junior work.

Get more SA tech and business news and subscribe to The Open Letter.