TRUE 2.0 — rebuilt around evidence.

The first version of TRUE did what most AI finance tools do: it answered confidently. It was fluent, it was fast, and it was — we eventually admitted — impossible for a user to verify. That's a fine property for a chatbot and a terrible one for a research tool.

Every answer now shows its work

A brief has four parts, in this order: the short read, the data TRUE actually checked, what remains uncertain, and the sources. The uncertainty section is not optional and it is not buried at the bottom in grey text. If the evidence doesn't support a confident answer, the answer isn't confident.

We made the depth visible

TRUE connects AI research with non-custodial trading, agents and copy trading. Your wallet signs app transactions; activated agents operate within the permissions you grant.

We added claim check

The most common way people lose money is acting on a confident claim that nobody checked. So we built a feature that does nothing but check claims — what the data supports, what it doesn't, and what the post conveniently omitted.

What we removed

This article describes an earlier product release. For current research, trading, agents and MCP workflows, use the Features directory and current setup guides. Historical release copy is not a statement of today’s account permissions or supported tools.

Some of that was a product decision and some of it was a values decision, and we're comfortable with both. The category is full of tools promising certainty in a domain that does not contain any. We'd rather build the thing that makes you a better analyst than the thing that sells more subscriptions.

Research is the product. The decision stays with you.