Why Anthropic Won’t Release This AI

Show notes

On this episode of Admin Access, Dominic and Evans are joined by Gabriel to break down some of the biggest moves happening in AI over the past few weeks The conversation kicks off with Anthropic's Claude Mythos, a model the company calls its most powerful yet and one it has pointedly refused to release to the public. Instead, it sits behind a restricted partner programs called Project Glasswing, accessible only to roughly 40 organizations, almost all hyper-scalers and enterprise giants from the US and Asia. The trio dig into what that really means. Is this genuine caution around a tool capable of autonomously finding and exploiting zero-day vulnerabilities? Or is it a carefully engineered piece of rage-bait marketing designed to dominate the conversation without shipping a single line of code to the public? And beneath all of that, where does Africa fit in a framework that's already quietly drawing the lines of who gets protected and who gets left exposed? With MTN moving money across 16 markets and M-Pesa serving millions across East Africa, the stakes of being locked out of these conversations are anything but abstract. They also share how they actually use AI coding tools in their day-to-day work, Claude Code, Codex, Gemini and rate their real experiences without the marketing gloss. The consensus lands somewhere between a 4 and a 7 depending on the task, the context, and frankly, the mood of the model that day. Powerful as a tool. Frustrating as a colleague. The episode wraps with Claude Design steps, a tool Dominic has been using . From moving from a rough spec to a high-fidelity prototype in minutes and what that shift means for UI/UX designers who haven't yet recalibrated their expectations. Learn more about our guest story here :https://www.youtube.com/watch?v=g2z0JQ2w-lw 00:00:Intro

Why Anthropic Won’t Release This AI

Show notes

In this episode