Watch a number get smaller
one move: descend the loss
Open the code pane immediately, then use the Spec tab, optional design question, tests, and artifact when you want them.
Finish each move before carrying the artifact forward.
A gradient points uphill, so subtract it to minimize a loss.
Deliverable: Write the mechanism in one sentence.Minimize f(x)=x² from x=5 with a fixed learning rate.
Deliverable: Type the smallest runnable implementation.The loss must decrease and x must approach zero.
Deliverable: Record the expected output and one edge case.Try a tiny rate, an oversized rate, and a second variable.
Deliverable: Change one variable and explain the result.A printed loss curve and a reusable gradient-descent loop.
Deliverable: Save the code, output, and a short failure note.The loss must decrease and x must approach zero.
Try a tiny rate, an oversized rate, and a second variable.
A printed loss curve and a reusable gradient-descent loop.
Mapped from the supplied Build Everything PDF, source page 4. The five moves are Deriva’s implementation contract for this project.