This page is deliberately small. We are a young company, and most impact pages are written years ahead of the impact. Ours lists what we are actually accountable for today — and a standing ledger you can check us against.
[ 01 ]AI responsibility
Y0 is an assistant that acts on a person's working life, so we hold it to instrument standards, not chatbot standards. Actions with consequences — sending, billing, committing — run on deterministic paths the user can predict. Reasoning is routed, traced, and signed: the runtime can always show why it knows what it knows.
Consent comes first, structurally. The trust kernel sits between every model and every byte of user context, and permissions are explicit, scoped, and revocable. We do not ship autonomy we cannot audit, and we do not call a system safe because nothing has gone wrong yet — the full architecture is published at /trust.
[ 02 ]Data ethics
Every read is attributable
The trust kernel permits a data access only when it traces to a task the user initiated. A granted scope is a loan, never a standing license — and the ledger records what was borrowed.
Consent is scoped, revocable, and legible
Permissions are granted narrowly, withdrawn in one action, and explained in one sentence each. If we cannot explain a read in one sentence, the feature waits until we can.
No training on user content without explicit opt-in
User documents, calendars, and tasks are context for that user's runtime, not raw material for ours. Any future exception requires a separate, specific, uncoerced yes.
No sale of data. No advertising use. No brokers.
Our revenue is the subscription. The moment user data funds the company through any other door, every promise above becomes negotiable — so the door stays bricked.
The user can take their context and leave
Export is a feature, not a support ticket. A trust ledger only means something if the auditor can close the account.
[ 03 ]Compute & environment
The runtime routes each task to the smallest model that clears its quality bar, and the read budgets that protect privacy also cap compute: a feature that reads a hundred things to produce one is rejected twice over. Efficiency is not our sustainability program — it is how the product is engineered, because we pay for every token and so does the planet.
On environmental claims, our posture is we measure before we claim. We do not yet have audited emissions data, so we publish none — no offsets purchased for the announcement, no net-zero pledge with a comfortable deadline. When we have numbers worth standing behind, they will appear in the ledger below, defined the way our KPIs are.
[ the test we apply ]
If a claim on this page cannot be checked,
it does not belong on this page.
[ 04 ]Community
Y0 was specified against the working lives of independent professionals here — designers, consultants, kirana-scale businesses — and the product is priced so that the people it was modeled on can afford it. Building for the Indian everyday first is a constraint that makes the product better everywhere, and we intend to keep it.
What we owe the builder community that trained us, we plan to repay in kind: our research notes are public, our teardown methods are documented, and as the team grows we intend to teach — runtime engineering, trust accounting, spec writing — to builders here, in person. That is intent, not yet record, and it is labelled accordingly.
[ 05 ]Transparency
Updated whenever a number changes. A zero here is a record we report, not a guarantee we sell.
✓ last updated 2026-06-11 — bengaluru, ist