AI systems rarely
fail loudly
They return fluent, well-formatted answers that are wrong. I build research and auditing software designed to catch that before you read it: several models answer independently, a synthesis step reconciles them, and it is allowed to conclude there is no consensus.
N° 01 — VenturesA small portfolio of
A small portfolio of
ambitious things.
What I’m building, and where each one stands.
N° 02 — MethodBuilt to prove
Built to prove
its own output.
I build with a system I wrote myself. It runs research, auditing and analysis — and its main job is establishing whether its own findings are actually true.
N° 03 — Contact
Get in touch.
If something here is useful to you, write to me directly.
contact@nstarlabs.com




