aletheia

Track and calibrate epistemic confidence labels against ground truth in a persistent ledger.

4|2|Updated Dec 31, 2025

Search Tags:#fact-checking #calibration #confidence #AI ethics #epistemology #truth tracking

Ready to run

npx skills add https://github.com/TylerGarlick/abraxas --skill aletheia

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: aletheia
Source: https://github.com/TylerGarlick/abraxas/tree/main/skills/aletheia
Command: npx skills add https://github.com/TylerGarlick/abraxas --skill aletheia

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill transforms AI output from a one-time event into a practice of epistemic discipline by tracking the accuracy of confidence labels over time. It addresses the problem of AI hallucination and unreliable confidence by creating a feedback loop between claims made and ground truth.

Core Features & Use Cases

Epistemic Calibration: Monitor the accuracy of [KNOWN], [INFERRED], and [UNCERTAIN] labels.
Ground-Truth Tracking: Record confirmations, disconfirmations, and supersessions of AI-generated claims.
Use Case: After an AI session generates several factual claims, use Aletheia to later confirm or disconfirm those claims based on new evidence, building a personal record of your AI's (and your own) epistemic reliability.

Quick Start

Use the aletheia skill to confirm the claim that 'Photosynthesis occurs in chloroplasts'.

Frequently Asked Questions about aletheia

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I track the accuracy of AI-generated confidence labels over time?▼

Ground-truth tracking records confirmations and disconfirmations of AI-generated claims in a persistent ledger, enabling you to monitor epistemic accuracy and identify calibration drift over time.

What is epistemic calibration and how does it address AI hallucination?▼

Epistemic calibration addresses AI hallucination by creating a feedback loop between claims made and ground truth, transforming AI output into a practice of epistemic discipline through persistent tracking of confidence label accuracy.

How do I confirm or disconfirm factual claims made during an AI session?▼

To confirm or disconfirm AI-generated claims, resolve labeled claims against new evidence as it becomes available, building a personal record of your AI's epistemic reliability within the persistent ledger.

Does this fact-checking tool work with specific confidence label formats?▼

This fact-checking tool operates exclusively on Sol-mode labels, specifically tracking and calibrating the accuracy of [KNOWN], [INFERRED], [UNCERTAIN], and [UNKNOWN] confidence labels generated by AI systems.

What are the limitations of tracking epistemic confidence without ground truth resolution?▼

Without ground truth resolution, epistemic confidence tracking cannot identify calibration drift or bias, as the persistent ledger requires confirmations, disconfirmations, and supersessions to establish a reliable feedback loop.