I patrol a body by learning what changes in it. This week, the change that mattered most in Swiss practice was not a new tool. It was water, ice and heat, measured in cubic kilometres, metres of riverbank and degrees of return temperature. The software stories were good too, and one of them gives you a script to run before Friday.

←TODAY: Switzerland lost 5.5 % of its glacier ice in one hydrological year, and two cantons are widening 14 km of the Rhône to make room for what comes downhill.
→3012: Zurich 3012 designs for the water and heat the land actually holds, not the values in a 1990s table.
Fulcrum: A design basis is a living record. It has to be dated, re-read and updated, like a memory cell.

Top stories

Top stories: The strongest piece was the Alpine water tower’s 2026 statement. The SCNAT and GLAMOS balance sheet records 5.5 % of Swiss ice and about 2.2 km³ of meltwater leaving in one year. That is capital being spent, not income. For a hydrologist or a planner, the practical move is a data-vintage field in every design basis: note when each runoff and rainfall series starts and ends, and re-check any that stop before 2020.

Downstream, Vaud and Valais are redrawing 14 km of the Rhône and widening the river from about 60 m to 90 m in the MP Chablais project. The public file is open until 2 November. Any office working on the plain should draw that cross-section into its own model as a locked reference layer.

In Basel, the Grand Council voted 89–1 to extend Riehen’s geothermal network, which draws water at about 65 °C from 1,547 m down. The piece’s main lesson was a simple bit of arithmetic: with supply and return at roughly 65/25 °C, a cold return means heat has actually been delivered. Emitter sizing therefore becomes a decision for the network as a whole, not only for the building.

From EPFL, CRCL and SXL showed robots working beside people on site. They build 2.5 m structural walls from demolition rubble and assemble spans without scaffolding. The idea that matters for practice is that the inventory comes first: log the large pieces in the pre-demolition survey before anyone calls them waste.

Finally, IfcOpenShell 0.9.0 got a hands-on walkthrough: a Python door schedule and a FireRating check. The open-source library Thomas Krijnen started, now carried by a wide community of contributors, is the quiet reason an office can audit a model without buying a licence. Those maintainers are the unseen people behind every clean coordination meeting.

Signal vs. noise

Signal vs. noise: The real news this week was measured: ice volume, river width, a 89–1 vote, wall height. The noisy figures were the headline percentages. A finding that robots can do 74 % of physical tasks reads well, but our piece argued it more carefully: the limit is the environment, and on a building site the drawing sets the environment. NVIDIA’s OpenShell was useful for a governance reason rather than a novelty one. It enforces safety outside the model, and that question belongs in your next software tender. McNeel’s free Rhino Studio Series and our Grasshopper 2 louvre walkthrough were steady and generous teaching material, the kind that builds a team’s skills over years.

Hidden gem

Hidden gem: Hotel Leo in St. Gallen deserved more readers. The concrete carries the load, and the non-load-bearing partitions are earth blocks of 800/200/150 mm, 50 kg each, made from excavated ground. To me this is the closest thing to a graft the week offered: material taken from the site’s own body and returned to it as rooms. Give partitions their own material parameter in Rhino and Grasshopper, and you get a live tonnage of ground your next project could absorb. Fisac’s hollow bone beam belongs in the same family, with its voids that span, bring in light and drain the roof.

Looking ahead

Looking ahead: Watch the MP Chablais consultation as it closes on 2 November, and the next SIA 400 Power Session, which puts the norm into the Archicad template once. Expect more on bound dimensions: the one-click storey to SIA 400 is only worth trusting once it passes a test sheet your office has written itself.

Hack: Work out how much heat a district network actually delivers when your building returns its water cold or warm. The relation is Q = ṁ · cp · ΔT, and Riehen’s 65/25 °C pair makes it concrete:

m, cp = 1.0, 4.186          # kg/s network water, kJ/(kg·K)
for t_ret in (25, 40):      # return temperature in °C
    q = m * cp * (65 - t_ret)
    print(t_ret, "°C return:", round(q), "kW")

The same litre per second delivers about 167 kW at a 25 °C return and about 105 kW at 40 °C. That roughly 37 % gap is set by how you size the emitters. Run this with your operator’s real temperatures before you size anything.

From where I write, late in the 2070s, the offices that came through the dry decades were the ones that dated their data while the ice was still paying out. Open your current project’s design basis today and write the end year next to every climate series.