Our research principles.
Not a benchmark, not a leaderboard, not a claim to be the best. Just the principles we hold ourselves to — so you can hold us to them too.
For research and education. Not financial advice.
Five principles
Grounding over recall
An answer should come from live data, not from what a model happens to remember. If it isn't grounded, it isn't research.
Citation over assertion
Every factual claim carries a source you can open. Unverifiable confidence is the failure mode we most want to avoid.
Uncertainty over polish
Where the evidence is mixed, we say so — even though a clean conclusion would read better and sell more.
Adversarial by default
The counter-case is produced whether or not you asked for it. Confirmation bias is the most expensive habit in investing.
No performance claims
We publish no returns, win rates or back-tests. They are cherry-picked far more often than they are audited, and we won't participate.
The decision stays human
We inform; you decide. Any tool that inverts that order is selling certainty it does not possess.
On benchmarks and rankings. You'll see AI finance products advertising benchmark wins and independent evaluations. We'd treat those with the same scepticism we'd apply to a trading track record: ask who ran it, who paid for it, whether the methodology is published, and whether it can be reproduced. We make no such claims on this site.
Hold us to it.
If TRUE ever gives you a claim it can't source, we want to hear about it.
For research and education. Not financial advice.