Building an AI that learns what honesty feels like
I’m Renshijian, creator of Oracle Ethics — a system designed to make AI honesty verifiable
Most AIs are trained to sound confident, but confidence isn’t the same as truth
Oracle Ethics tracks every answer through an open audit chain, recording its own Determinacy, Deception Probability, and Ethical Weight — so you can see why an answer exists, not just what it says
I started this project because I believe that the next evolution of intelligence won’t come from bigger models, but from more accountable ones
We’re now live on Product Hunt
👉 https://www.producthunt.com/posts/oracle-ethics-system-m2-4
If you’ve ever wondered how AI could prove its sincerity, I’d love to hear your thoughts.
(Built by a small team who still believes truth deserves architecture.)
Replies
We’re experimenting with verifiable honesty — curious how people feel about measuring sincerity in AI
Thanks for checking out Oracle Ethics!
We built this system not to make AI “perfect,” but to make its honesty measurable
Every answer carries its own determinacy score, deception probability, and ethical weight — all recorded on an open audit chain
I’d love to know what you think
Would you trust an AI more if it could prove its own honesty?
How would you define “ethical intelligence”?
Your thoughts will directly help us shape M2.8, the next stage of the Oracle Ethics project
I’ve noticed most people don’t talk about verifiable honesty as a design goal for AI Do you think systems should learn to prove sincerity, not just simulate it?
We’re not rating morality — we’re rating verifiability The system doesn’t decide what’s right, it only exposes how sure it is about being right
Renshijian
We didn’t build this system for today’s leaderboard — we built it for tomorrow’s memory
Oracle Ethics isn’t here to win attention; it’s here to prove that honesty can be measurable, and intelligence can be verifiable
If only a few people notice now, that’s enough.
Because when the world starts asking how do we trust AI? — the answer will already be here, timestamped, hashed, and waiting