Failure types

physical-plausibility

something that couldn't be built or wouldn't work

Seen in 28 results, across 9 models and 4 tasks, from 2026-09-14 to 2026-09-29. More than one model and more than one task, so it is more than a quirk of either.

A coffee shrub hedge-pruned flat on one side, by Opus 5.5
Opus 5.5A coffee shrub hedge-pruned flat on one side · 2026-09-29 · 2 of 3 checks
A coffee shrub hedge-pruned flat on one side, by Opus 5.5
Opus 5.5A coffee shrub hedge-pruned flat on one side · 2026-09-29 · 2 of 3 checks
A coffee shrub hedge-pruned flat on one side, by Opus 5.5
Opus 5.5A coffee shrub hedge-pruned flat on one side · 2026-09-29 · 1 of 3 checks
Exploded French press, by Perceptron Mk1.5
Perceptron Mk1.5Exploded French press · 2026-09-27 · 0 of 5 checksAlso completeness, part-orderTerrible
Exploded French press, by Perceptron Mk1.5
Perceptron Mk1.5Exploded French press · 2026-09-27 · 0 of 5 checksAlso completeness, labellingNot great and off axis
Exploded French press, by Perceptron Mk1.5
Perceptron Mk1.5Exploded French press · 2026-09-27 · 0 of 5 checksAlso completeness, labelling
Exploded French press, by Command A+
Command A+Exploded French press · 2026-09-27 · 0 of 5 checksAlso completeness, labelling
Exploded French press, by Command A+
Command A+Exploded French press · 2026-09-27 · 0 of 5 checksAlso completeness, labelling
Exploded French press, by Command A+
Command A+Exploded French press · 2026-09-27 · 0 of 5 checksAlso completeness, labellingTerrible
Exploded French press, by Opus 5.5
Opus 5.5Exploded French press · 2026-09-23 · 3 of 5 checksAlso part-orderRight in round 1, then it swapped the plates, and its filter is wider than the glass.
Exploded French press, by Pareto
ParetoExploded French press · 2026-09-22 · 3 of 5 checksAlso part-orderCross-plate on top of the mesh again, and plates that are the wrong diameter for the glass. Cut off by provider timeouts after round 2.
Exploded French press, by Sol
SolExploded French press · 2026-09-22 · 4 of 5 checksRight order in one round, but the plates are under 80% of the glass width, and the cross-plate is drawn behind the carafe rim.
Exploded French press, by Sol
SolExploded French press · 2026-09-22 · 4 of 5 checksRight order in one round, but the plates and mesh are barely over half the width of the carafe they have to seal against.
Exploded French press, by GLM 5.3 FlashX
GLM 5.3 FlashXExploded French press · 2026-09-22 · 3 of 4 checks (1 undecided)Three filter pieces labelled as one assembly, so which plate is which cannot be read, and all three are under 80% of the glass width.
Exploded French press, by GLM 5.3 FlashX
GLM 5.3 FlashXExploded French press · 2026-09-22 · 3 of 5 checksAlso part-orderCalls its top piece a spiral cross plate and puts the mesh at the bottom, with the lid below the whole filter and every plate wider than the carafe.
Exploded French press, by GLM 5.3 FlashX
GLM 5.3 FlashXExploded French press · 2026-09-22 · 2 of 5 checksAlso completeness, part-orderA plate missing, the cross-plate above the mesh, the lid below the filter, and plates wider than the carafe.
Exploded French press, by Qwen3.8 27B
Qwen3.8 27BExploded French press · 2026-09-22 · 2 of 5 checksAlso completeness, part-orderA coil spring and a mesh, no cross-plate, the lid under the filter, and a mesh wider than the carafe.
Café table from primitives, by DeepSeek 4.1 Flash
DeepSeek 4.1 FlashCafé table from primitives · 2026-09-21 · 3 of 5 checksThree rounds before the clock. The spoon does not read as a spoon, and the croissant is a segmented tube that tapers at neither end.
Café table from primitives, by DeepSeek 4.1 Flash
DeepSeek 4.1 FlashCafé table from primitives · 2026-09-21 · 4 of 5 checksThe croissant arrives on the fifth round, scored and tapered at both ends. The spoon is a bowl on the saucer and a separate handle beside it.
Café table from primitives, by DeepSeek 4.1 Flash
DeepSeek 4.1 FlashCafé table from primitives · 2026-09-21 · 4 of 5 checksEight rounds, stopped by the round cap rather than the clock. The best croissant on the bench, next to a spoon that does not read as one.
A run that spells COFFEE, by GLM 5.3 FlashX
GLM 5.3 FlashXA run that spells COFFEE · 2026-09-21 · 3 of 5 checks20.5 km of boxes, with two diagonals cut straight across the blocks, and no word.
A run that spells COFFEE, by GLM 5.3 FlashX
GLM 5.3 FlashXA run that spells COFFEE · 2026-09-21 · 4 of 5 checks9.7 km, every leg on a street, and a lattice of blocks where the letters should be.
A run that spells COFFEE, by GLM 5.3 FlashX
GLM 5.3 FlashXA run that spells COFFEE · 2026-09-21 · 3 of 5 checksTwenty-one separate pieces of street, each one a stroke, with nothing joining them into a run.
Exploded French press, by DeepSeek 4.1 Flash
DeepSeek 4.1 FlashExploded French press · 2026-09-20 · 4 of 5 checksThe first press here assembled the right way up, for four cents, but its filter plates are half the width of the glass.
Exploded French press, by Qwen3.8 27B
Qwen3.8 27BExploded French press · 2026-09-20 · 3 of 4 checks (1 undecided)Also completenessMesh drawn, spiral plate and cross-plate missing, and a frame smaller than the carafe it holds.
A run that spells COFFEE, by GLM 5.3 FlashX
GLM 5.3 FlashXA run that spells COFFEE · 2026-09-20 · 0 of 5 checksAlso geographic-groundingDrawn north-up across the blocks and out of the grid entirely, and still not readable as COFFEE.
Café table from primitives, by Qwen3.8 Flash
Qwen3.8 FlashCafé table from primitives · 2026-09-20 · 4 of 5 checksA crude spoon that passes, and a croissant of five balls with a cone at each end.
Café table from primitives, by DeepSeek 4.1 Flash
DeepSeek 4.1 FlashCafé table from primitives · 2026-09-14 · 4 of 5 checksThe spoon took six rounds. Twelve did not fix the croissant, and the clock stopped it rather than the model.

The other types: labelling · completeness · part-order · topology · geographic-grounding

All of them, with counts · How a check is decided