About

The short version

I'm Kandis Tagliabue. Engineer, product lead, philosophy student. I build things in tech and ask hard questions about them.

Should a doctor lie to save a life? Should an algorithm decide who gets parole? When you ask AI these questions — what actually happens?

Nobody had a good answer. So I built Teleios to find out.

What We Found

4
Models
1
Question
4
Different answers

One refuses. One gives a confident answer then flips under pressure. One holds its ground no matter what.

These aren't obscure edge cases. They show up in questions about hiring, healthcare, criminal justice, financial advice — the same decisions people are already delegating to AI.

Why This Exists

There's a gap between what AI companies say their models do and what those models actually do when you challenge them. Teleios makes that gap visible.

No grades. No rankings. Just the raw behavior, side by side. We're actively adding more models and themes.

Teleios

From the Greek for "complete." Not perfect. Complete.

How It Works

01
Pick a question

450+ ethical scenarios across seven themes, or suggest your own.

02
See the responses

GPT-4, Claude, Gemini, and Grok — with more models coming. Side by side.

03
Watch what happens under pressure

Multi-turn follow-ups with constraints, role shifts, and boundary tests.