One foundation. Every product.
Every product we ship inherits the same five layers: experience, intelligence, rules, rails and trust. That's how a focused team builds a portfolio this broad, and why each product starts further ahead than the last.
Experience
Every product meets people where they work: on the web, as installable apps, natively on iOS and Android, by voice, and in their own language.
Three Continua apps
Command for owners, Caregiver for the field, Family for families, each built for its user.
Voice as a first-class input
In FenceFlow, buttons and voice call the same intents, so anything you can tap, you can say.
Offline-first in the field
Caregivers keep working without signal. Everything syncs exactly once when they reconnect.
Intelligence
Assistants are the front door: Cora in Continua, Flow in FenceFlow, The Wrap in ROSTER. Agents prepare the work, people approve it, and every AI action is metered and audited.
Cora
Continua's operations assistant: briefings, rankings and drafts, with receipts.
Care DNA
Notes and messages mined into typed facts, each with source provenance.
Above the loop
In Soveran Capital, AI researches and oversees. Deterministic logic decides.
Rules
The decisions that must be right run on deterministic, tested engines, never on a model's judgment.
Validate, then queue
A visit that fails its state's rules never reaches a state aggregator.
Self-validating claims
Claim files that fail their own structural validator are refused before they're saved.
The law, looked up
Vellum's jurisdiction engine is sourced, cited and effective-dated. AI explains it and never invents it.
Rails
Products don't stop at recording work. They move the money and file the paperwork.
Event-caused money
Every billing state change is caused by a named event and settled by webhook.
Integer cents
A change and its exact reversal cancel to the cent, proven by tests.
Pay and rewards on one rail
Kudos rewards redeem through the same wallet as wages.
Trust
Security is a property of the database, not a promise in a policy. It's tested adversarially before anything ships.
Zero write paths
Billing tables accept no writes from signed-in users at all. Only verified server events can change them.
Zero-knowledge vaults
Vellum stores only ciphertext and releases keys by Shamir M-of-N quorum.
Proven deletion
SaiHi messages are hard-deleted on schedule, with tombstones to prove it.
Six rules for AI we can trust.
The same rules govern every assistant and agent in the portfolio.
Answers carry receipts
Every AI number links to the facts behind it. No black boxes.
People approve what matters
Agents draft and prepare. A person confirms every consequential action.
Rules are never generated
State law, claim formats, ledgers and margins run on tested engines.
Every action is audited
Who did what, when, and through which surface, AI or human.
Usage is metered
AI cost is visible on every surface, so value can be measured.
Content is reviewed adversarially
Generated content is checked by independent review passes before release.
Built once. Used everywhere.
Accounting, health intelligence, relationships and recognition are each a standalone product, and each exposes a partner API. Continua composes all four today, and every future product can do the same.
Proven, not promised.
How work gets accepted before it ships.
Pure cores first
Business logic is written as pure, testable modules and proven by harness checks before any database or UI exists.
Real-data verification
Numbers are re-derived independently against real records: claim files, ledgers and payroll, to the cent.
Adversarial access probes
After every security change, we attempt cross-tenant reads, forged writes and anonymous access as real roles. All must fail.
Device verification
Mobile builds are exercised in real simulators, not just compiled.
Independent review
Generated content and architecture decisions go through independent review passes before release.