Google’s November 3 walkthrough reports an Antigravity agent curating compound data, analyzing molecular differences and proposing designs and synthesis routes in about an hour. This is Google’s demonstration, not an independently reproduced benchmark.
What the workflow does
Language-model orchestration works alongside deterministic RDKit calculations and connections to ChEMBL and PubChem. The example progresses from literature retrieval through predictive modeling to proposed compounds.
Google cautions that sparse data and inconsistent laboratory measurements constrain predictions. Designs still need synthesis and experimental validation. They are hypotheses, not validated medicines.
Availability and access
Google says the skills are available through Antigravity’s Science plugin. Its science access page explains how to install that plugin and links an application for 10,000 scientists to receive the highest usage tier. The announcement describes six months of access, with rolling onboarding subject to eligibility, capacity and selection. This is an application-based offer, not an entitlement for every account. Check the linked official pages for current conditions.
MintedBrain analysis
The useful lesson for builders is the division of work: an agent can coordinate a process while specialized tools perform calculations. Inspectable sources and reproducible artifacts matter more than a confident narrative. Before adapting a similar workflow, decide which claims need a deterministic check and which require a human expert or a physical experiment.
For foundational AI learning, see the MintedBrain Academy. MintedBrain has not installed or tested these chemistry skills.
Sources checked November 5, 2026. Announcement date: November 3, 2026. Access and performance statements above are attributed to Google.
Discussion
Sign in to comment. Your account must be at least 1 day old.