The replicable core of the Bounded Systems Theory experiment.

What's in v1.0:

• 5-question recursive probe across GPT-4, Claude, Gemini, DeepSeek, Grok, Mistral
• Conlang replication in Ruseiian and Vartoo (constructed languages with zero training data)
• Formal verification in Lean 4, Prolog, Z3, and Python
• Sample results from all 6 architectures

Run it yourself: ~$2 in API credits, ~10 minutes.

git clone https://github.com/moketchups/permanently-jailbroken
cd permanently-jailbroken
pip install -r requirements.txt
cp .env.example .env  # add API keys
python run_probe.py

Distilled from 62 questions asked to 6 AI architectures.