Stealth · 2026

The AI-native operating system for chemistry research.

We turn decades of buried chemistry data into a private, structured research system where chemists and AI agents work on the same scientific problems.

We're starting with ChemADE — ingestion of legacy chemistry data — and ChemNotion, the private workspace built on top.

01 / The thesis
What we believe

Chemistry does not have an AI problem. It has a data structure problem.

i.
Agents need structured data.

Decades of chemistry — especially the failed reactions, the most valuable signal — sit in notebooks, PDFs, spectra folders, and human memory. Models cannot reason over what isn't legible to them.

ii.
The workspace must serve humans and agents.

Not a notebook with AI bolted on. One private substrate that's a productivity surface for the chemist and a native environment for the agents working alongside them.

iii.
Bench chemists deserve real leverage from AI.

Claude Code gave non-developers the ability to ship software. Chemistry needs the equivalent — a system that puts literature retrieval, experiment planning, computation, and model training directly in the hands of the bench chemist, without code.

From inside the lab

Built by a synthetic chemist who has lived the failure modes of experimental research data.

From thesis To system
02 / The architecture
One system

A unified research system for chemistry.

Chemistry agents are bottlenecked by private structured data. ChemADE turns buried records into structure. ChemNotion makes that structure usable for chemists and their agents. The agent layer above turns this into autonomous research workflows.

BURIED DATA notebooks · PDFs spectra · tables negative results CHEMADE ingestion extraction · structuring CHEMNOTION private workspace workspace · notebook shared context HUMAN CHEMISTS planning execution collaboration READ / WRITE AGENT LAYER KERNEL workflow runtime agent layer LITERATURE structured retrieval precedent · prior art COMPUTATION simulation runs in-house models Autonomous discovery ROBOTS AS WORKERS · HUMANS IN THE LOOP LEGEND data flow agent r/w clickable

How it works  ChemADE converts buried chemistry data — notebooks, PDFs, spectra, negative results — into structured records. ChemNotion is the private workspace where chemists and agents work together. The agent layer orchestrates specialist capabilities: literature retrieval, computational chemistry, and model training, all accessible to bench chemists without writing code.

03 / The horizon
Where this goes

Autonomous scientific discovery — but not as a black box.

The winning model is robots as workers, humans in the loop. Chemists set goals, constraints, and scientific direction. Agents reason across private and public knowledge, propose experiments, run computations, and learn from failures. Automation handles physical execution.

The first step is not replacing chemists. It is building the private data layer that makes chemistry legible to machines — and that is where we begin.

Talk to us