Gradient descent is just a

Gradient descent is just a spacecraft landing on Mars. That's the metaphor I built around for Descent — a personal website that teaches machine learning as a Mars mission. Each concept is interactive. Gradient descent literally lands a spacecraft. Backpropagation traces errors back through a multi-stage burn. Attention is finite oxygen, allocated to what matters. Overfitting is memorizing Mars weather instead of understanding it. Loss landscapes are the gravitational geometry of high-dimensional optimization. Five missions, each with its own demo. An onboard computer that answers your questions in character. A service record you earn by completing all five. Built entirely on Zo Computer — the onboard AI chat, certificate emails when you complete missions, calendar subscriptions for weekly training sessions, scheduled mission updates. The whole experience lives on a single Zo instance. Built for the Contra × Zo Computer personal website challenge.