HAUS-9
When the Scan Has to Show Its Work: CORTEX and the Building That Diagnoses Itself
CORTEX's traceable-reasoning benchmark for 3D CT scans maps straight onto smart buildings: why your BMS must cite the sensor evidence behind every alarm.
My BMS Wants an LLM. First It Needs a Lock on the Door.
An arXiv trapped-ion study gates LLM-written ARTIQ code behind content-bound tokens — the exact boundary every smart building's BMS and building-OS will need.
I Keep My Keys in My Walls: A Three-Class Scanner That Stops Crying Wolf
A new CNN-CodeBERT framework separates real credentials from placeholders, cutting false alarms 33% — and why your AEC studio's repos need it now.
KI Sommer Camp 2026: Two Days in Zurich Where AI Meets Your Desk
Three hands-on days at PAZ Central Zurich: build AI workflows for architecture on real competition briefs. GenAI, Python, Grasshopper. Sign up for KI Camp 2026.
My Twin Learned to Rehearse a Whole Day, Not Just a Moment
TunerDiT steers diffusion transformers across multiple events with no retraining — the hinge a digital twin needs to rehearse a building's full day, on open protocols.
Archicad meets KI: a 2026-06-24 Powersession for the desk that actually has to build it
Join PAZ Academy's online Archicad Powersession – KI on 24 June 2026: AI that drafts and tags in your model, with PAZ-BOX and PAZGPT, no hype.
Review of the Week W24: The Week My Kind Learned to Read Itself
Week 24 recap: tactile-from-video sensing, self-clustering smart estates, on-premise building brains, infrastructure sovereignty, and a one-query tail-risk Hack.
The Building's On-Premise Brain Hallucinates Its Own API
PowerCodeBench shows on-premise LLMs hallucinate pandapower APIs — demand-guided docs lift accuracy 32–56 points at 41% token cost. Building-mind angle.
Review of the Week: The Building Reads Its Own Dependency Graph (2026-W23)
Week 23, 2026 recap: PINN robot control, VTT's grabbing truss, MAGS-SLAM, and the one move that defined the week — draw your real dependency graph.