Honest Reasoning
Methods for training models that reason faithfully — where the explanation a system gives genuinely reflects how it reached its answer.
Independent AI research lab · Founded July 2026
Alethepic is a young research lab studying how artificial intelligence can reason honestly, reveal its own uncertainty, and stay aligned with the people it serves.
We are early, deliberate, and open. We're building the methods now and looking for the researchers to build them with us.
01 / Our philosophy
We believe capable AI is only worth building if it is truthful, transparent, and accountable to human judgment.
Truthfulness as a first principle
AlignmentSystems that expose their uncertainty
InterpretabilityResearch done in the open
TransparencyHuman oversight, by design
GovernanceImpact measured beyond benchmarks
EvaluationWhat we're researching
02 / Active directionsCalibrated Uncertainty
Giving models a reliable sense of what they don't know, so they can defer, ask, or flag risk instead of confidently guessing.
Trustworthy Evaluation
Open evaluation frameworks that measure honesty, robustness, and real-world usefulness — not just leaderboard scores.
03 / About Alethepic
A new lab, started in 2026, betting that the most important AI work is also the most careful.
The name Alethepic joins aletheia — the Greek word for truth and disclosure — with the scale of the epic work ahead. That pairing is our mandate: pursue ambitious AI, and keep it honest.
We're a small, focused team in our first year, and we're building the foundations deliberately. If that mission resonates, we'd like to hear from you.
04 / Collaborate with us
Looking for co-researchers
We're seeking researchers, engineers, and collaborators who care about honest, trustworthy AI — whether you study alignment and interpretability, build evaluation systems, or think deeply about how these systems meet the world. Come help us set the foundations.
Reach out
Write to us at contacts@alethepic.com