Dishaya / Principles

How We Decide When Nobody Is Watching.

Most values pages are decoration. This one is the public version of the constitution Dishaya is actually run on: seven principles, stated plainly enough that you can catch us breaking them.

The rules, the one disclosed exception, and how to hold us to all of it.

What This Page Is

These seven principles decide what Dishaya builds, what it charges, and what it refuses to do. They are commitments, not marketing. When one of them conflicts with a feature or a growth idea, the principle wins.

1. Honesty Over Fluency

AI is rewarded for sounding right. We are not. We would rather show you less and be trusted than show you more and be doubted, so every claim in a Dishaya report carries one of four labels: verified, partial, unverified, or contradicted. The uncomfortable labels are published, not filtered out.

"Contradicted" is the label that costs us the most to show, and it is the one that matters most. It means the cited source says the opposite of the claim, and you learned that from us instead of from the person you sent the work to. A tool that only ever says "verified" is a rubber stamp. We are not building a rubber stamp.

2. Your Content Is Never Ours

We do not train on your content. Not now, not by a quiet policy update later, not in any form. Your chats and documents live on your device, and what you research is your business, not our raw material.

There is exactly one exception, and we disclose it instead of burying it: research reports are stored with your account so your Library can reopen them. They are deleted when your account is deleted, and they are never mined, profiled, or used for anything except serving them back to you. The full posture is on our trust page.

3. Evidence Over Confidence

A confident tone is not evidence. In Dishaya, every claim is checked against its exact source passage before it is written, and every claim carries both its verdict and its source, one tap away. Confidence is shown, never implied.

Each report closes with the Confidence Ledger: a plain tally of how the evidence held up, what passed, what failed the check, and what the run cost. If we cannot support a sentence, we tell you so instead of hoping you never look. What the labels do and do not promise is spelled out in our methodology.

4. Honest Pricing

You see the cost of every run, before you commit to it and after it finishes. There are no dark patterns anywhere in the product: no hidden fees, no surprise charges, no renewal tricks, no cancellation maze. Cancelling works the first time you try, from inside the product, without a phone call.

The free tier is honest too. It is a real, usable product, not a demo engineered to frustrate you into paying. If we cannot afford to offer something, we say so plainly rather than quietly making it worse.

5. Calm Over Clever

Software that demands your attention is easy to build. Software that lowers your heart rate is not, and that is the bar the interface is held to: fewer options, better defaults, no badges begging to be clicked, nothing blinking for its own sake.

Every screen should make the next step obvious. When a control would make the product look more impressive but make the person using it more anxious, the control loses.

6. Long-Term Thinking

Dishaya is built to be trusted in ten years, not to trend for ten days. That rules out growth that costs trust: no attention-harvesting funnels, no selling what we know about you, no slow erosion of promises once enough people depend on us.

It also means commitments with dates attached. Our methodology page carries a standing commitment to report our accuracy publicly every quarter, including the quarters where the numbers are unflattering. A principle you cannot be audited against is a slogan.

7. Responsible AI

Dishaya is a rigor tool, not an oracle. The labels tell you how well each claim is supported by the source it cites; a source can itself be wrong, and we never claim more than the labels measure. We use multiple AI models where appropriate, and the promise we make is about the output you can inspect, not about any machine being infallible.

Humans stay accountable for what they send. Dishaya shows you exactly where to verify, and the decision to stand behind the work remains yours. Dishaya is not legal, medical, or financial advice, and we will not pretend otherwise to close a sale.

Common Questions

Do these principles actually constrain what you build?

Yes. This page is the public version of the constitution the company is run on. When a feature, a growth idea, or a pricing change conflicts with one of these principles, the principle wins and the idea is dropped or rewritten. We publish them so you can hold us to them.

Do you train on my content?

No. Never, in any form. Your chats and documents live on your device. The one disclosed exception: research reports are stored with your account so your Library can reopen them. They are deleted with your account and are never mined or used for anything except serving them back to you.

What happens when you fall short?

We say so. Inside the product, what fails the check is disclosed, never hidden. The same rule applies to the company: mistakes are written up in the changelog in plain language, and our methodology page carries a standing commitment to report accuracy publicly every quarter, including quarters where the numbers are unflattering.

See The Principles In The Product

Start Free