spaceship (autonomous-ai) errors
11 recurring problems engineers report in spaceship (autonomous-ai), ranked by impact.
#41 by errors trackedMCADPLMVerification
Recurring spaceship (autonomous-ai) errors, by impact
- 1The assembly drawing generator's read-back passed a balloon on a part's far face hidden behind another part, and the new check requires an arrow to end on a visible outline.MCAD · Agent failure · 1 report74
- 2The lead flagged parts-list rows that give 'ISO 2768' as the tolerance, which breaks their rule that the drawing gives tolerances. A new check found 58 rows across 11 team lists, plus 18 copies in the built master BOM lists, 76 total.PLM · Agent failure · 1 report74
- 3The master BOM build only listed a to-do when a team row used a different name for a register-owned number, and a reviewer showed putting an old part name back still gave a green build, with 24 team rows now carrying old register names.PLM · Agent failure · 1 report74
- 4The release checklist used its own out-of-date copy of the mechanical check's reader, taking 'X' from the size in a fill-piece name as the noun, so WALL's 20 fill-piece lines missed the fill rule.PLM · Agent failure · 1 report74
- 5Assembly mode never counted a quantity given in sets, so WALL's cassette clips listed as 518003, '4 set' stayed one placement across four clip zones and the read-back failed.PLM · Agent failure · 1 report66
- 6Fusion crashed twice on Nov 1 right after fetching the team folder's second page of files, costing every team about 20 minutes each time, and the fix is a 10 s pause after a new file.MCAD · Agent failure · 1 report66
- 7On a pod build, PR #353 moves holes to match PEM nut and earth-stud datasheets, marks two sizes and positions as assumptions, and supersedes the drawing for the fourth time.Verification · Agent failure · 1 report66
- 8On a pod build, two seal-line gaps were found by eye on Nov 1 before anyone had a check for them. PR #352 traces the seal as closed rings from each part STEP faces at 91 cut heights.Verification · Agent failure · 1 report66
- 9Six A1 joint sheets for a modular enclosure: the ceiling closing seam's rivnuts are not in the CEILING STEP, so they are drawn from the joint register's sizes and marked 'not in CAD'. A coverage check finds sheets for only 6 of 16 register joints on main, and several sheets wait on other PRs.Verification · Agent failure · 1 report66
- 10Six design rulings and 23 new part numbers are logged after checking main, open PRs, recent branches and chats, while three old numbers can only be cancelled once AIR's list drops them, since a cancelled number still in a list fails the BOM check.PLM · Agent failure · 1 report66
- 11The weldment cut-list reader could not read two-cut ends or a note like '(to within 0.1 deg)', so 10 of 63 frame members failed on the text alone, and a part used in four rings printed one ring's name.MCAD · Agent failure · 1 report66
How it fails
Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.