Summary
Creator Magic’s Mike Russell pays out of pocket to put Kimi K3 — the new frontier coding model from Moonshot AI — through its paces against Claude with Fable Five, after K3 jumped to the top of the frontend code leaderboard the day of its release. The model ships with 2.8 trillion parameters, a one-million-token context window, and open weights promised later in the month, and is accessible now through Moonshot’s $19/month entry plan via the Kimi Code CLI in YOLO mode.
Russell runs four parallel tasks simultaneously in Kimi Code and Claude Code: a playable voxel sandbox game, a motion-graphics product launch video rendered entirely in Python, a neon synth pad with drum sequencing, and a website for his Creator Magic brand built with live web search. Results are genuinely mixed. Kimi K3’s voxel game arguably edges out Fable Five on visual complexity, and both synthesizers are impressively functional. But K3 burns through a $19 plan in minutes with 403 and 404 errors repeatedly halting all four projects, requiring additional top-ups just to see results — a significant reliability concern for paid users.
The video is a useful real-world stress test of a model that leaderboard numbers alone can’t capture. It documents where K3 delivers on its promise (output quality on creative coding tasks) and where the current Moonshot AI platform falls short (service reliability, rate limits, cost predictability) — making it practical reading for any developer evaluating whether to migrate workloads from Claude or GPT-based tooling.
📺 Source: Creator Magic · Published July 17, 2026
🏷️ Format: Comparison
