Truth you can verify.
Most AI confidently makes things up. Cosmo is truthful by construction: sourced facts, an honest "I don’t know," and advice it only gives once it has verified it can act.
It is the founding property of an AI Operating System: memory is sourced, or it does not exist.
Three rules that make Cosmo honest.
Not slogans about trust, rules built into Cosmo that govern every fact it keeps and every answer it gives.
Sourced, or dropped
Every fact Cosmo holds about you is backed by a verbatim quote from your own data, an email, a document, a calendar entry. If a claim can't be sourced, it never enters Cosmo's memory. No inference dressed up as fact.
See the source quote on every fact, in the Mirror.
Knows what it doesn't know
When the answer isn’t in your data, Cosmo says so, plainly. It abstains instead of guessing, and tells you exactly what it would need to know for sure. An honest "I don’t know" beats a confident wrong answer every time.
Abstention is a feature, not a failure.
Honest advice, verify-gated
Before Cosmo acts, it checks that it can actually do what it just claimed. Everything Cosmo offers you is honesty-verified against a real, certified capability, so it never promises a thing it cannot deliver.
Tagged by provenance: live, memory, or unverified.
A chatbot guesses. Cosmo cites.
Capabilities we certify, not just claim.
An autonomous QA loop tests every capability and certifies it on a renewing TTL. What you see below is what Cosmo will actually do, tagged with its status and the time it was last verified. We publish ours; competitors hide theirs.
A record we publish, not a promise behind the curtain.
Truth is verifiable only when it is published: every capability wears its status and its last-verified time where you can see it, not in an internal report you never read. When a bug is found, the capability is marked with it and Cosmo will not silently rely on it. We show our whole record; others just claim.
Frequently asked
What does "truthful by construction" mean?
Every fact Cosmo holds about you is backed by a verbatim quote from your own data, an email, a document, a calendar entry. If a claim cannot be sourced, it never enters Cosmo's memory. Truthfulness is not a mode you switch on; it is how the AI Operating System is built.
How do I know Cosmo is not hallucinating?
When the answer is not in your data, Cosmo abstains and says so plainly instead of guessing. Every claim is tagged by provenance (live, memory, or unverified) and every fact in the Mirror carries the source quote it came from, which you can read, correct, or forget.
Are the capabilities OIKON lists actually verified?
Yes. An autonomous QA loop tests every capability and certifies it on a renewing freshness window. The /quality page publishes the live status of each capability (certified, in certification, bug open, or disabled) with the time it was last verified. We publish ours; competitors hide theirs.
What happens when a capability has a bug?
A known issue is logged and the capability is marked bug-open or disabled, and Cosmo will not silently rely on it. Before Cosmo acts, every suggestion is honesty-verified against a real, certified capability, so it never promises something it cannot deliver.
- Not just an AI.
- Not just an agent.
- Not just an operator.
- Not just a pocket buddy.
Cosmo is all of them, and more.
Truthfulness isn’t a mode you turn on. It is how Cosmo is built, across every surface, every capability, every plan.
Trust an AI that proves it.
Connect your email and watch Cosmo build a source-cited picture of your week, every fact backed by your own data. Free, on Fable 5.