7 April 2026: The Date That Made AI Feel Different
Most Number of the Day episodes begin with a figure that lands somewhere in South African life almost immediately. A price. A percentage. A loss. A gain.
This one begins with a date.
And that matters.
Because 7 April 2026 is not important for what it measures. It is important for what it signalled. In this episode, Gareth Edwards and Francis Herd unpack the moment Anthropic warned that Claude Mythos, one of its newest AI models, was too powerful for public release. That is not the kind of sentence people are used to hearing from the companies building this technology. It sounds less like a product update and more like the start of a new chapter.
The heart of the concern is cybersecurity. Anthropic says Mythos has reached a level of capability where it can surpass all but the most skilled humans at finding and
exploiting software vulnerabilities. It says the model has already found thousands of high-severity vulnerabilities, including some in every major operating system and web browser. That changes the mood instantly. This is no longer just a conversation about productivity, convenience, or clever prompts. This is a conversation about critical systems and what happens when machines become exceptionally good at breaking them.
That is why Gareth and Francis instinctively bring the story back to infrastructure. Water. Electricity. Servers. Networks. The invisible digital scaffolding holding up ordinary life. This is where the episode gets its weight. Not because it chases the most apocalyptic version of the story, but because it understands that modern risk now lives inside software. If software can be compromised faster, deeper, and more autonomously than before, then the consequences are not abstract. They are social, economic, and national. Cisco describes this moment as a critical inflection point, where AI is giving defenders huge new capabilities while also lowering the threshold for attackers. That tension runs through the whole episode.
And yet the story is not simply that AI is dangerous. It is that AI is becoming double-edged in a way that feels newly difficult to manage. Anthropic is not publicly releasing Mythos, but it is giving access to selected partners through Project Glasswing so they can secure critical software and help patch vulnerabilities. Anthropic says more than 40 additional organisations responsible for critical software infrastructure are also being brought into that effort. So the same technology raising alarm is also being framed as part of the defence. That contradiction is not a side note. It is the story.
The smartest move the episode makes is bringing all of this back to South Africa at the end. Not through panic, but through familiarity. AI is not only arriving as a geopolitical cyber-risk story. It is also arriving as a shopping assistant, a recommendation tool, a digital convenience folded quietly into daily life. That contrast is what gives the conversation its sting. The future is not arriving in one dramatic shape. It is arriving in fragments.
Some of those fragments feel helpful. Some feel ominous.
And 7 April 2026 may be the date when a lot more people start realising both things can be true at once.
Catch up on all Number of the Day episodes here: https://www.enca.com/number-day-podcast