Watch a neuron learn to output 1.0
one move: train by reducing 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.
Forward, loss, chain rule, and update form one complete learning loop.
Deliverable: Write the mechanism in one sentence.Train weights and bias so a ReLU neuron reaches a target without PyTorch.
Deliverable: Type the smallest runnable implementation.Loss must fall and the final prediction must approach the target.
Deliverable: Record the expected output and one edge case.Change the target, learning rate, activation, and add a second neuron.
Deliverable: Change one variable and explain the result.A hand-trained neuron with its loss trace.
Deliverable: Save the code, output, and a short failure note.Loss must fall and the final prediction must approach the target.
Change the target, learning rate, activation, and add a second neuron.
A hand-trained neuron with its loss trace.
Mapped from the supplied Build Everything PDF, source page 6. The five moves are Deriva’s implementation contract for this project.