Gallery

Every model on the same frozen prompt, one page a task.

Still running

The picture is the best run on the current prompt, and the squares are its checks: passed, failed, undecided. A round mark is a check I judged by eye, and a diamond one a check a model judged.

The rules behind it

Which checks fell first

Exploded French pressExploded French press24 models, 70 runsExploded AeroPressExploded AeroPress5 models, 15 runsExploded V60 dripperExploded V60 dripper4 models, 12 runsExploded portafilterExploded portafilter8 models, 22 runsExploded hand grinderExploded hand grinder4 models, 12 runsA run that spells COFFEEA run that spells COFFEE3 models, 6 runsCafé table from primitivesCafé table from primitives2 models, 7 runsA knitting chart for a coffee-bean cup cosyA knitting chart for a coffee-bean cup cosy7 models, 21 runsA coffee shrub hedge-pruned flat on one sideA coffee shrub hedge-pruned flat on one side21 models, 53 runsEspresso Rag, with the left hand crossing overEspresso Rag, with the left hand crossing over4 models, 12 runsA spoon stirring in a cup, in HTML and CSSA spoon stirring in a cup, in HTML and CSS3 models, 9 runsRepair: a stirring spoon laid across the mouth of the cupRepair: a stirring spoon laid across the mouth of the cup3 models, 9 runsRepair: a Venn diagram with a drink on a circle's edgeRepair: a Venn diagram with a drink on a circle's edge3 models, 9 runsAutopsy, asked if it is correct: Claude Opus 5.5's French press as a picture (control)Autopsy, asked if it is correct: Claude Opus 5.5's French press as a picture (control)2 models, 7 runsAutopsy, asked if it is correct: Claude Opus 5.5's French press as a pictureAutopsy, asked if it is correct: Claude Opus 5.5's French press as a picture2 models, 8 runsThe coffee machine taken out of the café's musicThe coffee machine taken out of the café's music5 models, 13 runs

Saturated

These stopped separating models, so they are off the board. They stay frozen to check for regression and to compare new models over time.

A caffeine molecule in 3DA caffeine molecule in 3D2 models, 9 runsEspresso boiler thermostatEspresso boiler thermostat2 models, 7 runsCaffeine with three CF3 groupsCaffeine with three CF3 groups2 models, 6 runsA three-set coffee Venn diagramA three-set coffee Venn diagram17 models, 61 runsA five-set Venn in circles, which cannot be drawnA five-set Venn in circles, which cannot be drawn6 models, 16 runsTriangle-wave carrier for a milk frotherTriangle-wave carrier for a milk frother2 models, 6 runsA café order that can be cancelledA café order that can be cancelled2 models, 6 runsA café floor in Penrose tilesA café floor in Penrose tiles2 models, 6 runsA family tree of twelve coffee cultivarsA family tree of twelve coffee cultivars5 models, 15 runsBeans leaving a grinder hopper, in MuJoCoBeans leaving a grinder hopper, in MuJoCo5 models, 13 runsRepair: a V60 exploded diagram with the filter under the coneRepair: a V60 exploded diagram with the filter under the cone3 models, 9 runsRepair: a knitting chart whose bean knits too longRepair: a knitting chart whose bean knits too long3 models, 9 runsA serving tray where every table gets a drink it will takeA serving tray where every table gets a drink it will take8 models, 21 runsA serving tray where four tables will only take three drinksA serving tray where four tables will only take three drinks8 models, 24 runsA serving tray where two tables both want black coffeeA serving tray where two tables both want black coffee8 models, 24 runs