What you can trust on this site, and why

The research engine (its agents, AI models, databases and controls) runs privately. It pushes a small set of public-safe records to this site every few seconds through a publication gateway. This site cannot reach back into the engine.

• Visitors cannot start agents, trigger simulations, change hypotheses or budgets, pause the engine or send prompts.
• Loading a page never triggers an AI call or a simulation; many visitors cost the research engine nothing.
• If the engine goes offline, this site keeps working, shows the last published state, and says clearly when it last heard from the engine.

Published within seconds: agent activity, hand-offs between agents, papers processed, simulations started and completed, and daily counts.

Published after human review: findings and experiment recommendations.

Never published: AI prompts or hidden reasoning, internal logs and errors, API keys, file paths, machine or network details, operator identity, and payment or personal details.

Every public record passes a fixed schema that rejects unknown fields, and an automatic scanner that withholds anything resembling a secret, a path, an address or an error trace, twice (before sending and on arrival).

• Payments are handled on Stripe's (or PayPal's) secure pages. FoltzTech never sees or stores card numbers.
• A payment counts only after the provider's cryptographically signed notification arrives; the browser cannot add money to the totals.
• Refunds and chargebacks are subtracted. Test payments are never mixed with real ones.
• Spending is shown by category and only when it is actually recorded. See funding.
• Contributions are not tax-deductible unless this site says otherwise after FoltzTech obtains nonprofit status.

• Plain-English summaries are written by rule-based code that restates results; the wording is careful but can be formulaic.
• Human counterparts of yeast genes come from a small curated table that is still being re-verified.
• Research-area tags are keyword rules, so a few may be approximate.
• The Digital Cell animates one representative parameter set; the spread across model variants is summarised as agreement percentages.
• Nothing on this site has reached human clinical evidence.