Lila Sciences' AI Lab Screens 2,942 Catalysts in 3 Months, Reviving a Discarded Path to Green Hydrogen

Lila Sciences has published results from its autonomous science platform on arXiv: an AI-guided automated lab proposed, synthesized and screened 2,942 catalysts in 3 months, identifying 6 high-performing material families for the oxygen evolution reaction (OER) in proton-exchange-membrane electrolyzers. Among them, an iridium/ruthenium-free palladium-based oxide (such as the InMnPdOx system) — which John Gregoire, a scientist with 12 years studying the reaction, initially deemed certainly wrong — ran stably for over 1,000 hours under harsh acidic conditions with almost no performance degradation, becoming the strongest candidate. Kenneth Stanley shared and praised the result.

Confirmed

• 2,942 catalysts proposed, synthesized and screened in 3 months; 6 high-performing material families identified
• The palladium-based catalyst contains no iridium or ruthenium and ran stably for over 1,000 hours in acidic OER conditions with negligible degradation
• The direction had been abandoned by the field; John Gregoire initially believed the AI's formulation must be wrong
• The team released a paper on arXiv disclosing specific catalyst systems such as InMnPdOx
• The platform exceeds 90% automation, integrating combinatorial sputtering synthesis, high-throughput screening and ML composition-property modeling
• Kenneth Stanley shared and liked the result

Why it matters

• Standard OER catalysts rely on iridium oxide; iridium is among the rarest metals on Earth with only a few tonnes mined annually, constraining green-hydrogen scale-up
• The case shows autonomous AI labs can recover material directions excluded by expert intuition, accelerating clean-energy materials R&D
• The highly automated closed-loop platform (synthesis—screening—modeling) offers a replicable paradigm for materials discovery

2026-09-25 ~ 2026-09-26 · 8 related posts

• [source] Lila Sciences' AI lab screens 2,942 catalysts in 3 months, cracks green hydrogen's OER problem — tw_killian · 2026-09-25
• [source] Lila Sciences' AI finds palladium catalyst a 12-year expert rejected, running 1,000+ hours — AndrewLBeam · 2026-09-25
• AI-proposed catalyst a 12-year expert called wrong becomes strongest clean-hydrogen lead, lasts 1,000+ hours — AndrewLBeam · 2026-09-25
• AI-proposed catalyst scoffed at by 12-year expert runs 1,000+ hours in clean hydrogen test — tw_killian · 2026-09-26
• AI proposed chemistry a 12-year expert called wrong — it became their best hydrogen catalyst — tw_killian · 2026-09-26
• [source] AI-driven lab finds iridium-free palladium catalyst InMnPdOx stable for 1,000 hours — CatAstro_Piyush · 2026-09-26

2 near-duplicate retellings: AndrewLBeam · CatAstro_Piyush