A GPU, a notebook,
no setup.

Six kinds of practice surface. Half are guided, the rest are quiet. All run in your browser, on real hardware, with the tutor a click away.

№ 01The six surfaces

Guided notebooks

Step-by-step interactive notebooks. Each cell has a checkpoint, each checkpoint has a hint.

Free sandbox

GPU

An open environment with GPU access. Build whatever you want, share the link.

Challenge labs

GPU

Timed problem-solving. You get a starter, a test suite, and a clock.

Pair coding

GPU

Real-time collaboration with a peer or a coach. Learn together, build together.

API playground

Compare LLMs side-by-side. Same prompt, different models, plain diffs.

Competition arena

GPU

ML competitions with a leaderboard. Win a printed certificate; lose, and we tell you why.

№ 02A notebook, mid-flight

transformer_from_scratch.ipynb

GPU · A10GPython 3.11

Cell · 01

import torch
import torch.nn as nn

class MultiHeadAttention(nn.Module):
    def __init__(self, d_model, n_heads):
        super().__init__()
        self.d_k = d_model // n_heads
        self.W_q = nn.Linear(d_model, d_model)
        self.W_k = nn.Linear(d_model, d_model)
        self.W_v = nn.Linear(d_model, d_model)

Shift + Enter

Output

MultiHeadAttention(
  (W_q): Linear(512, 512)
  (W_k): Linear(512, 512)
  (W_v): Linear(512, 512)
)
Parameters: 786,432
GPU Memory: 3.2 MB

The coach says

Attention dimensions look right. Next: implement forward() with scaled dot-product.

Still cooking

More surfaces on the way.

Challenges, competitions, and a small dataset library are in development. We'll add them when they're ready, not before.

A GPU, a notebook,no setup.