A lab robot can run all night on a synthesis that was doomed from the start. A night of robot time and thousands in reagents, spent on a recipe that was never going to work. That waste is not rare. Berkeley's A-Lab reported a 71% success rate, meaning about 29% of its attempts never yielded the target (A-Lab, Nature, 2023). A lot of that is impossibility you could catch first: a charge that doesn't balance, a tin(II) mix that oxidizes in humid air. So in building our Autonomous Lab AI demo, we put a deterministic gate in front of the robots. Before anything is synthesized, plain code (not the AI) checks it against five published chemistry rules: Goldschmidt tolerance, octahedral factor, charge neutrality, Sn(II) oxidation hazard, and the organic-precursor decomposition ceiling. Feasible passes. Doomed gets blocked, with the violated rule and the reagents and robot-hours it spared. On one recorded campaign against a synthetic benchmark, the gate blocked 38 doomed syntheses before they ran, saving $4,419 and 90.1 robot-hours, with zero infeasible runs executed. Every decision was written to a hash-chained ALCOA+ record. Edit one entry and the chain visibly breaks. Agents advise, code decides. When your lab runs itself overnight, what stops a doomed experiment before it burns the reagents, and can you prove afterward why each one ran? #LabAutomation #SelfDrivingLabs #MaterialsDiscovery #LabInformatics #RDautomation