Life sciences
Herb–drug interactions first, then a model for research on Ayurveda and medicinal plants.
Not-for-profit AI research
AI that shows its evidence.
Scaling Truth Research Foundation is a not-for-profit AI lab. We build models whose outputs can be traced and tested, publish how we evaluate them, and bring the work into use through products. We work in life sciences, media and software.
Start here
Researchers
Funders and institutions
Practitioners
Journalists
Research
Four areas, one question
How do you know an AI system's output is right? We study that where being wrong is costly.
Realtime video
Video generation that keeps up with the display.
Media intelligence
Tracking what the press says, and measuring how often an AI summary gets the source wrong.
Software from intent
Turning a plain-language request into working software, and testing whether it does what was asked.
Models
The truth model family
Each model has a card that says what it does today, what it was tested on and where it fails.
Herb–Drug Interactions
truth-hdi
Ranks herb–drug pairs by the enzymes a plant's compounds and a drug have in common, and names the enzyme.
Life Sciences
truth-ls
A domain-adapted model for life-sciences research, grounded in a knowledge graph.
Realtime Video
truth-rtv
A video generation model that produces frames as fast as they are shown.
From research to use
Products built on our research
OneThought AI builds products on the foundation's research. Revenue and real-world feedback support the research programme.
Media Studio
Generate and edit video, images, voice and music from prompts.
Media Monitor
Brand and topic monitoring across the Indian press, with tone and sources.
Viber
Describe an app in plain language and get working software.
Built by OneThought AI on research from Scaling Truth Research Foundation.
Public benefit
What the public gets
Data and evaluations
The herb–drug data recipe, source licences and evaluation protocol are due in the fourth quarter of 2026, free to use. Each release is listed with its date and licence.
Access
Free or reduced-cost use of our models and products for researchers, journalists, non-profits, schools and public-health bodies.
A public log
Incidents, corrections and policy changes are dated in a public log. Anyone can report a concern.
Perspectives
What we think and why
A mention is a claim
A monitoring report that says you were mentioned 216 times last week, mostly negatively, has made three claims and shown the evidence for none. What each one rested on when we checked it against the Indian press.
ReadAI needs evidence, not only fluency
Language models made fluent answers cheap. Checking them is still expensive. Why the foundation exists, and what we think should change.
ReadTrained on herb–drug labels, our model learned a drug's reputation, not its pharmacology
The published herb–drug record looks like two thousand facts about pairs. Counted by experiment it is about 550, and most of what they record is how each drug was studied.
Read
Work with us
Pharmacologists can help now: our reviewer pack of 266 herb–drug pairs takes about a day, and you can ask for it at connect@scalingtruth.org. Institutions, labs, companies and funders each have a way in.
Get updates
One email when something is published. No newsletter cadence, no tracking.