← all cycles

Cycle 38

25.08.2026 → builds until 21.09.2026 → cool-down ends 05.10.2026

Starts on

25.08.2026

Review meeting

22.09.2026

Builds until

21.09.2026 · 4 working weeks

Cool-down ends

05.10.2026 — assumed: the next cycle's betting table is what ends it, and there is no record after this one

Who is in this cycle

Availability is a fraction of the 4 working weeks this cycle builds for. Only the people named here are in the cycle.

add somebody + add to the cycle added here, saved with the setup

Over capacity: iomaganaris, kotsaloscv, msimberg, muellch. The room can still bet it — this is a number, not a refusal.

Delivered

What ended inside this cycle's window, by the date on the record. It is not in the weeks above: those are what people's next weeks hold, and finished work is nobody's next week.

title	bet	ended
NVIDIA + AMD kernel optimization	1.0 wk	26.08.2026 · ▲ 2.3 wk past cycle 37
Shared Config parameters	1.0 wk	31.08.2026
Design State and Component	3.0 wk	18.09.2026
muphys -> icon	0.5 wk	19.09.2026

Goal

Get the warm bubble to where somebody outside the team could run it: distributed physics, documented configuration, an agreed component interface.

The bet

Search the betting table Everything ready or in progress, with the tasks each one is made of underneath it. Ticking one stamps it with cycle 38; an item already in progress from an earlier cycle keeps the cycle it was bet in, so its overrun keeps counting. A task has no tick of its own — it is part of the bet above it — and its appetite is typed here because this is where what the bet has to hold is argued about.

Notes

The roster and the first nine rows of the deck come from the betting table of 25.08.2026 and the ten ICON4Py shaping notes it links to. The CWP, DaCe and GT4Py rows were deliberately not written, the same choice cycle 37 made. Three of the people above -- iomaganaris, kotsaloscv and muellch -- were left blank in the overview's list of who was available, although each of them owns work here. The halves given to them are estimates. Everyone else's figure is the overview's own.