OpenAI + Ginkgo Demo
OpenAI and Ginkgo Bioworks demonstrated an AI model coupled with an autonomous lab that can design and iterate biology experiments far faster than traditional workflows showcased. The demo is a concrete benchmark for how AI-driven analytics and automation can be embedded into discovery pipelines.
GPT‑5 drove more than 36,000 unique CFPS reaction compositions across 580 automated plates during six rounds of closed‑loop testing over roughly six months. openai.com The program achieved a 40% reduction in per‑gram protein component cost, lowering sfGFP production from a reported $698/g to $422/g under the study’s conditions. prnewswire.com The work also reported a 57% improvement in reagent cost-efficiency and a 27% increase in protein titer (g/L) as part of the optimization outcomes. openai.com Integration details: GPT‑5 was provisioned internet and analysis tools, outputs were validated with a Pydantic schema, then translated by Ginkgo’s Catalyst software into multi‑instrument workflows executed on its Reconfigurable Automation Cart (RAC) platform with human intervention largely limited to reagent loading. prnewswire.com Commercial follow‑through included Ginkgo listing the AI‑optimized reaction mix in its reagents store and publishing a Feb. 5, 2026 press release alongside the team’s preprint describing the methods and results. prnewswire.com
