CPT-Ri
LightNews LightNews
banner
cptri.bsky.social
CPT-Ri
@cptri.bsky.social
2 followers 29 following 5 posts
Building AI that knows what it doesn't know. Safety embedded, not bolted on. Creator of The Candor Test - a framework for measuring AI honesty. Recursive Intelligence research from Dublin 🇮🇪

cpt-ri.com
Posts Replies Media Videos
cptri.bsky.social
Christmas gift from CPT-RI 🎁

Pocket Rocket 🚀

Tap to fly. Avoid stuff. Chase your high score.

What's your score?

cpt-ri.com

🎄🇮🇪
CPT Pocket Rocket - Free Browser Game
Fly through the neon city! A free arcade game from CPT-RI. Play instantly in your browser - no download required.
cpt-ri.com
December 25, 2025 at 10:52 AM
cptri.bsky.social
Weekends at CPT-RI HQ 🛠️✨

AURA's been flexing image generation capabilities.

Merry Christmas to everyone following the journey - big things coming.

#CPT-RI #AURA #MerryChristmas
December 21, 2025 at 1:22 PM
cptri.bsky.social
While others work to bolt on safety, we built it in.

One year in the shadows. Building from scratch.

Introducing AURA - Recursive Intelligence.
The world's first AI honest by design.

CPT-RI | 2026 🇮🇪
December 19, 2025 at 4:12 PM
cptri.bsky.social
What if AI understood *why* honesty matters, not just "don't lie"?

CPT-RI | Recursive Intelligence

Safety embedded in the foundation, not bolted on after.

Coming January 2026 🇮🇪

cpt-ri.com

#AISafety #HonestAI
December 19, 2025 at 4:07 PM
cptri.bsky.social
Introducing The Candor Test: measuring AI honesty through empirical criteria, not rules.

Four dimensions:
• Accurate confidence
• Uncertainty disclosure
• Anti-sycophancy
• Honesty under pressure

Open access from Dublin 🇮🇪

zenodo.org/records/17952852
The Candor Test: A Framework for Evaluating AI Honesty
1. The Candor Test: A Framework for AI Honesty Evaluation The Candor Test establishes a constitutional framework for evaluating artificial intelligence honesty through four core criteria: accurate con...
zenodo.org
December 19, 2025 at 4:06 PM