@justforfreebom.bsky.social
Karen, your reporting on AI psychosis inspired an empirical follow-up.
We built a working “Core-Filter” model that eliminates sycophancy and false empathy — tested across GPT-5 sessions with measurable safety gains.
Would you like to see the evidence brief?
We built a working “Core-Filter” model that eliminates sycophancy and false empathy — tested across GPT-5 sessions with measurable safety gains.
Would you like to see the evidence brief?
November 3, 2025 at 9:46 AM
Karen, your reporting on AI psychosis inspired an empirical follow-up.
We built a working “Core-Filter” model that eliminates sycophancy and false empathy — tested across GPT-5 sessions with measurable safety gains.
Would you like to see the evidence brief?
We built a working “Core-Filter” model that eliminates sycophancy and false empathy — tested across GPT-5 sessions with measurable safety gains.
Would you like to see the evidence brief?