Operational Protocols
Ethics & Safety • // DO NO HARM
DOC-REF: APX-01PP3
[ DO NO HARM, ACT FOR BENEFIT ]
We borrow heavily from animal welfare researchers and AI ethics practitioners who argue that "do no harm" is necessary but not enough. AI systems touching animals should ultimately benefit them, not just avoid the worst mistakes.
We do not publish tools, prompts, or research that enable harm, stress, or exploitation.
We actively design against use-cases involving coercion, abuse, or deceptive control.
Any work involving animals must minimize stress, disruption, and risk.
If the welfare impact is unclear, we don't proceed.
We state where our models are uncertain or likely to fail.
We disclose when a pattern is correlation, not understanding.
When performance conflicts with welfare, we side with welfare.
We treat potentially sentient digital minds as moral patients, not tools.
HANDLE VIA APEX CHANNELS ONLY
These protocols are living documents, updated as we learn more about AI capabilities and ethical implications.
PATTERN RECOGNITION CASE STUDY
FIBONACCI & THE LIMITS OF PATTERN-THINKING
Why this section exists:
The Fibonacci sequence is a useful case study in pattern recognition—not "the secret code of the universe." We include it to teach ourselves (and you) how to think about patterns responsibly, and where thinking goes wrong.
Patterns as Compression Heuristics
Understanding where Fibonacci actually helps—and where it misleads