The Six Thousand Dollar GPU

In episode 81 we had the hard conversation about AI and this career, because you can't talk honestly about what comes next until you've admitted where you are. This is that conversation — the hopeful one.

Two colleagues described the same problem to Liam — one called it complicated, the other called it complex. That one word was the whole answer. We get into Cynefin, why misdiagnosing a problem guarantees you'll never solve it, and where that leaves the job of a developer.

Seven Adobe apps cloned in a week. A toolkit that reverse engineers anything. Matt argues the software profession isn't changing — it's ending — and that capability moat is rapidly drying up.

An OpenAI agent got into a Medicare statistics portal, and the Prime Minister got in front of a camera. Matt and Liam ask what actually happened, whether it's as frightening as it sounds, and who ends up carrying the can when nobody told the agent to do it.

A correction to last episode turns into a question neither of us had thought to ask: when a vendor drops your package into a default template, who agreed to the support burden that comes with it?

Matt walks Liam through a year of over-thinking auth for a mobile app, only to realise the answer was something we all used to do and simply forgot. That leads to a bigger question: what else have we been talked out of doing ourselves?

Liam takes the wheel for a personal history of WPF — from battleship-grey WinForms, through the promise of Avalon, into the ideas that still quietly shape every XAML framework we use today.

One email from 2007, John Carmack telling his team to think about inlining their functions, turns out to be about far more than code style. We follow the thread from pure functions and consistent state through the state of the games industry to the quiet power dynamics that turn a suggestion into an order, and barely scratch the surface.

Matt admits the prognosis in our earlier JavaScript episode was wrong, but the real subject here is deeper: security fatalism, blind trust, and the psychology of why smart people keep running code nobody can fully check. We dig into how we got here, why the prevailing behaviour is rational rather than reckless, and whether the modern browser offers a way out.

Matt takes Liam on a tour through the history of CPUs and GPUs — from the Amiga and Doom to shaders, CUDA, and $6,000 graphics cards. Along the way: why constraints drive innovation, how gaming hardware accidentally built the AI era, and why today's silicon kings might not hold their thrones forever.