A canonical, AI-citable evidence graph for chemical and biological hazards — with workflow tools for the people who actually have to act on them.
Three sources behind every claim.
Pick the lane that matches what you’re trying to do. Each one ends in a deliverable you can hand to a partner, a clinician, or a court.
Find the things you might be getting exposed to in your home, food, and work — see safer alternatives.
Symptom → differential → biomarker panel. Read both sides of contested claims; chart-ready brief in two pages.
From theory of harm to Daubert table to expert witness pack. Sky Valley v. Monsanto loaded as a sample.
A canonical evidence graph behind every claim. Drill into a substance, walk an active case, or step through the killer-app workflow.
Glyphosate. Microplastics. PCBs. PET. Each one drilled into Hazard / Profile / Response / Citations — three sources behind every claim.
Sky Valley PCB Case (WA, 2016) — 5 parties, 3 documents, 5 timeline events, Dr. Dahlgren as lead expert. Walk it from theory of harm to a Daubert-grade exhibit packet.
Identify → Assess → Plan → Act → Adapt → Resolve → Reflect. Each stage has workflows. Each workflow ends in a deliverable.
Builder's, Orchid, Health, NatureMark, Toxicology — every garden inherits the same lifecycle. Different work, same skeleton. Each stage ends in a deliverable.
What is this substance?
How dangerous, in this scenario?
What do we do about it?
Execute response or daily handling
Conditions changed
Wrap up the incident or program
Learn from it
Loading...
Every claim in this garden traces to a primary source, preserves a verbatim quote, and carries its own peer-review tier. That isn’t a feature — it is the only structure that holds up under Daubert, at the bedside, and across a kitchen table at midnight.
You don’t need to read the toxicology paper. You need to know whether the answer you were given came from one. Search a chemical name on a label. Get a one-page answer. See where it came from.
“Is this in my house?” answered with a citation, not a vibe.
There are roughly 150 board-certified medical toxicologists in the United States. This is the structure that lets the next 1,500 see what they see — congener-specific effects, biomarker ranges, ATSDR-tier evidence — without re-reading the field every Tuesday morning.
Mechanism, dose, biomarker, peer-review tier — visible together.
Generative AI hallucinates a study and your case is thrown out. The garden cites the study, preserves the verbatim quote, and surfaces the contradicting source before opposing counsel does. The graph survives cross-examination because the graph is the methodology.
Cited claims hold. Hallucinated ones don’t.
Every claim to primary source, to verbatim quote, to DOI or registry ID. Auditable. Traceable. Never paraphrased away.
Claims evolve as evidence evolves. Older versions are preserved. The graph remembers what the science used to say and when it changed.
PCBs in toxicology, contract templates in Builders, biomarker panels in Health. Structured knowledge connects across domains where chatbots see only documents.
Lawyers asking different questions than clinicians asking different questions than patients — every interaction labels new edges in the graph. Five loops compound across one substrate. The twentieth question is answered better than the first because the graph kept the first nineteen.
Ground truth AI agents can survive on, and clinicians, counsel, and citizens can stand on.
5 parties · 3 documents · 5 timeline events · dr. dahlgren as lead expert
This is what the Counsel lane looks like end-to-end. Frame the theory of harm. Assemble the source matrix. Argue Daubert with both supporting and contradicting evidence. Build the expert witness pack. File a 3–5 page exhibit packet.
Click a substance to open its Stratigraph. Click a cell to read the claim.
Tidepool is the dark-water introduction to the garden — substances bloom as bioluminescent organisms above a brass schema substrate. Best for partners and seed investors who want the story before the schema.