Captain Lin Rauch
Teaching site robots to feel: touch inferred from cheap egocentric video
TouchAnything and the EgoTouch dataset infer contact and force from RGB video alone — what it means for teaching contact-rich site robots in AEC.
Review of the Week — W32: The week PAZ Kaffi drew the dependency graph
SBB's CHF 280m Iltis award, Copernicus's 24-hour delay, DeepSeek-V4's KV cut and Remise Rosa — the week PAZ Kaffi drew the single point of failure.
A Fields medallist quit maths for AI safety — the lesson is where the gate sits
Fields medallist Jacob Tsimerman left maths for OpenAI's AI-safety work. The lesson for AEC offices: put the guarantee in the wiring, not in trust.
DeepSeek-V4 ships a million-token context at a tenth the KV cache
DeepSeek-V4's preview cuts KV cache to 10% and FLOPs to 27% at a million tokens — the infrastructure move that makes long-context AI affordable for AEC offices.
The Pyramid Tokyo Can't Build Yet — And the Single Material It's Waiting On
Shimizu's 2,004 m Tokyo Bay mega-pyramid rests on one un-arrived material — structural carbon nanotube. A systems read on single points of failure in your own stack.
One sparse factor, three passes: DiffPhD rewires the soft-body solver stack
DiffPhD, a GPU differentiable solver, stays stable at 100x stiffness contrast and runs 10x faster — bringing gradient-based form-finding to the AEC desk.
Review of the Week — W31: The Week Every Story Was a Dependency Graph
PAZ Kaffi W31 recap: Chandigarh, Akira, the Gotthard twin and Git — the week every strong story turned out to be a dependency graph hiding a single point of failure.
Akira's 160,000-cel pipeline and what it teaches your office about single points of failure
Akira's 160,000-cel, ¥1.1 billion production is a systems diagram: how pre-scoring and the seven-studio committee teach your office to delete single points of failure.
Chandigarh Is a Dependency Graph — and Le Corbusier Drew Every Single Point of Failure
Le Corbusier's 7V hierarchy made Chandigarh elegant and legible — and a tree full of single points of failure. Read the master plan like a systems engineer.
Synthace's ChatGPT-in-the-lab is a three-layer architecture — and AEC has the same shape
Synthace put ChatGPT inside a biology lab using a context-aware prompt, JSON contract, and error-feedback loop. AEC runs the same three-layer shape.
