Little Coder — video demos

Interactive visuals for the Little Coder video. Every page is a single self-contained HTML file — open it in a browser, no build step, no server. All support light and dark mode.

Terminal-Bench comparisonterminal-bench.html

Where Qwen 3.6 + Little Coder lands against Gemini CLI and the frontier. Animated bars with a replay button.

The two contendersmodel-comparison.html

Gemma 4 31B (dense, QAT) vs Qwen 3.6 35B-A3B (MoE) side by side — and why one feels 10× faster.

One name, many brainsmodel-family.html

The Gemma 4 / Qwen 3.6 family explorer — brain sizes, file sizes per quant, and what hardware each fits on.

Quantization, livequantization.html

The "photo over Messenger" analogy — watch an image degrade as bits drop, with real bits/weight and file sizes.

Can I run it?can-i-run-it.html

Pick hardware, model size, quant, and context — get memory fit, tokens/sec, and a verdict.

Running a local modelhosting-flow.html

Step-through pipeline: download → load into RAM → context window → localhost server → Little Coder.

The experiment

The agent promptprompt.html

The exact one-shot prompt both models were given to build the vacation-tracking SaaS, with a copy button.

Assets for the Little Coder episode · github.com/OverseedAI/little-coder-assets