{"stable_id":"8a2da4a32ef78159","slug":"berkeley-alab-autonomous-synthesis","company":"Lawrence Berkeley National Laboratory and University of California, Berkeley","workflow_name":"A-Lab closed-loop autonomous materials synthesis","function_code":"engineering","pattern_codes":["autonomous_with_backstop"],"changed_assumption":"That scientists must manually choose every recipe, execute every synthesis step, interpret each XRD result, and plan the next experiment.","evidence_strength":"verified","publication_tier":"showcase","freshness":"current","reviewed_at":"2026-08-23","updated_at":"2026-08-22","source_quality_summary":"1 peer reviewed; publication outcomes are verified.","caveat_summary":"Research demonstration rather than routine production; 21 targets were not realized and final patterns were manually refined.","freshness_caveat":null,"workflow_summary":{"before":[{"actor":"Materials scientist","action":"Select a target and design a synthesis recipe from literature and thermodynamics.","actor_type":"control"},{"actor":"Laboratory staff","action":"Dose, mix, heat, characterize, and interpret each experiment.","actor_type":"human"}],"hinge":"That scientists must manually choose every recipe, execute every synthesis step, interpret each XRD result.","after":[{"actor":"ML recipe system and robots","action":"Propose recipes and execute powder dosing, heating, and XRD characterization.","actor_type":"system"},{"actor":"ML analysis and active learning","action":"Assess phase yield and propose a new reaction path after failure.","actor_type":"system"},{"actor":"Scientists","action":"Select target space, audit outputs, manually refine final diffraction patterns, and diagnose system failure modes.","actor_type":"system"}],"decision_mode":"bounded_autonomy","decision_marker":"AI acts within a human backstop"},"before":[{"order":1,"actor":"Materials scientist","action":"Select a target and design a synthesis recipe from literature and thermodynamics.","handoff_to":"Laboratory staff","control":"Scientific judgment"},{"order":2,"actor":"Laboratory staff","action":"Dose, mix, heat, characterize, and interpret each experiment.","handoff_to":"Scientist","control":"Manual lab procedures"}],"after":[{"order":1,"actor":"ML recipe system and robots","action":"Propose recipes and execute powder dosing, heating, and XRD characterization.","handoff_to":"ML analysis","control":"Air-stable targets and robotic constraints"},{"order":2,"actor":"ML analysis and active learning","action":"Assess phase yield and propose a new reaction path after failure.","handoff_to":"Robotic laboratory","control":"Target-yield threshold"},{"order":3,"actor":"Scientists","action":"Select target space, audit outputs, manually refine final diffraction patterns, and diagnose system failure modes.","handoff_to":"Research record","control":"Human scientific accountability"}],"decision_rights":"The platform selects follow-up recipes inside the target set; scientists define targets and perform final validation.","exception_path":"Failed targets are inspected for synthesis or computational failure; scientists can adjust decision rules and manually validate diffraction patterns.","removed_work":["Manual execution of 353 repeated experiments","Manual selection of each follow-up recipe"],"outcomes":[{"metric":"Target compounds synthesized","baseline":"Manual, serial scientist-directed synthesis cycles","result":"36 of 57 target compounds synthesized, a 63% success rate","period":"17 days of continuous closed-loop operation","scale":"353 experiments spanning 33 elements and 40 structural prototypes","attribution_caveat":"Research demonstration rather than routine production; 21 targets were not realized and final patterns were manually refined.","evidence_label":"verified"}],"executive_lesson":"Autonomy came from closing the loop between prediction, physical execution, measurement, and replanning; the 37% miss rate shows why scientist validation remains part of the system.","anti_pattern":"Reporting only successful compounds or describing the lab as human-free.","questions_for_leaders":["Is measurement fast enough to close the loop?","Who adjudicates model and instrument disagreement?"],"collections":["embodied-work"],"bundle_version":"1.0.0","bundle_fingerprint":"sha256:c23c6cc2b88153d008ea8fda928f632ce0011fc2d4c5036672a16e5d895bab93","canonical_url":"https://brianletort.ai/transformations/berkeley-alab-autonomous-synthesis"}