Noor Kade
Thrust Network Analysis: The 350-Year-Old Theorem That Certifies Vaults and Grows New Shells
How Thrust Network Analysis turns Hooke's 1676 hanging chain into a lower-bound proof that certifies Gothic vaults and form-finds new compression shells.
SONG rehearses the crowd before the robot walks your lobby
SONG's 3D Gaussian splatting simulator trains social-navigation robots on photoreal crowds — and turns every public splat-twin into a Swiss data-residency question.
Reading the Platform: Why an SBB Node Is Switzerland's Highest-Value Urban Problem
An SBB node is never a station — it's this platform, this 800m walkshed. Read transit-oriented density from the timetable and rank your parcels first.
Own Your AI: Run Hugging Face Models Locally Before You Rent Them
A hands-on PAZ tutorial: install Hugging Face transformers, run inference and embeddings on your own hardware, and keep your studio's tender data on-site.
Obsidian: A Local-First Second Brain You Actually Own
Obsidian rebuilds Luhmann's Zettelkasten as plain Markdown files you own. A hands-on PAZ tutorial: clone, link, git-version your studio's knowledge graph.
Zurich's City Twin: Reading Heat and Density in the GIS Before a Building Moves
Zurich's open GIS layers plus live simulation form a city digital twin that reads heat, traffic and density before a building moves — with a PostGIS join to try.
GAF: the 4D robotics plugin your atelier should commission, not license
Gaussian Action Field extends 3D splatting into a 4D robotic-manipulation engine. Why Swiss ateliers should commission it, not license its wrapper.
When the AI Platform Goes Dark: Run It Locally So Your Work Survives
A free AI platform vanished overnight and 80 games nearly went with it. Learn to run models locally with transformers.js so your work outlives the vendor.
Seaborn: the quiet library that makes a column of numbers tell the truth
A foundation guide to Seaborn, the high-level Python library over matplotlib — how declarative statistical graphics make the honest chart the easy one to draw.
The blurry-mean trap: why your AV (and BIM) model can score well and still be wrong
An arXiv diffusion world model shows SSIM and cosine metrics reward useless blur — the evaluation lesson every AEC ML pipeline needs. Plus a 3-line Python check.
