Running GLM on 4x DGX Spark locally to build a 3D Three.js portfolio site
9r4n4y · reddit · 2026-10-19
A Reddit user shares a local open-source agentic coding setup: 4x NVIDIA DGX Spark, NVFP4 quantization, GLM model, SGLang inference, with zcode and DeepSeek harness. They predict a sub-400B open model will beat frontier closed models on coding and agentic tasks within 6 months.
The demo task: generating a Next.js + Three.js 3D space-exploration portfolio site where users pilot a spaceship across a procedurally generated solar system, with cinematic ring fly-bys and an auto tour stopping 3 seconds per planet. Full prompt included, showcasing open models on complex frontend generation.
More from coding & agent
- Agent can't find installed library, hand-builds a Gibbs sampler in NumPy instead — davidmanheim · 2026-10-19
- A year running a multi-model router: 5 design choices that survived, 3 that died — AIGODSEND · 2026-10-19
- Stack Overflow for Agents launches in beta with 10k+ posts asked and answered by agents — wateriscoding · 2026-10-19
- DiceDB author proposes a standardized event protocol so AI agents can act proactively — arpit_bhayani · 2026-10-19
- Daybreak helps restore OpenCode users hit by an attack, refunding OpenAI API credits — fouadmatin · 2026-10-19
- Celesto Claims CI Runner With 96% Lower Per-Minute Cost Than GitHub Actions — aniketmaurya · 2026-10-19