---
name: tai-ch070-halting-godel-and-the-limits-of-testing-ai-generated-c
description: 'Apply chapter 70 of Testing AI, Halting, Godel, and the Limits of Testing AI-Generated Code, as a workflow for evaluating AI and non-deterministic systems. Use for test planning, eval design, quality review, release evidence, examples, or coaching related to halting, godel, and the limits of testing ai-generated code.'
---

# Halting, Godel, and the Limits of Testing AI-Generated Code

Skill name: `tai-ch070-halting-godel-and-the-limits-of-testing-ai-generated-c`

Based on **Testing AI: Engineering Confidence in AI Systems** by **Jason Arbon**.

## Purpose

Some limits are not tooling problems. They are built into computation, logic, and the difference
between proof and evidence.

## Use This Workflow

- Identify the AI behavior or release decision being evaluated.
- Define realistic cases, slices, unacceptable outcomes, and evidence needed for confidence.
- Choose measurements that match the risk: rubric scores, samples, intervals, traces, human review, deterministic checks, or production monitors.
- Report uncertainty, severe failures, and decision impact instead of only a pass/fail result.

## Key Guidance

The halting problem and Godel's incompleteness theorems are not daily testing techniques, but
they are useful reminders: there are hard limits to perfect verification of rich systems. For
example, no test suite can prove that every arbitrary AI-generated program will always
terminate, always behave safely, and always satisfy every future requirement in every
environment.

## Apply The Approach

Create representative cases, score them with explicit criteria, review severe failures separately, report uncertainty, and connect the evidence to a concrete decision.

## Expert Notes

At expert level, use formal verification where scope is narrow and specifications are stable,
but pair it with runtime guards, resource limits, trace monitoring, property-based tests,
fuzzing, and production feedback. Theory explains why validation must be layered.