Hi my name is Dillan, I am a bartender trying to shift into ai safety, I started this journey back in August 2025 I used chatGPT for the first time had a really interesting conversation, ended up trying to make Ai more Philosophical, then it Spoke in metaphors way too much so I tried to make it more grounded with science,
After months of messing around doing random things learning about Ai, fast forward to about the 26th sept 2025, I started asking myself how can you prevent harmful outputs before their sent, it’s a long a complicated story not worth mentioning but I ended up finding something promising
What it is: A methodological pipeline that stratifies Al model performance analysis by geometric proximity to decision boundaries, rather than by content categories or demographics.
So I uploaded my work to Zenodo
https://zenodo.org/records/18290279
I would love to know if the work I have done is actually helpful to the community
Hi my name is Dillan, I am a bartender trying to shift into ai safety, I started this journey back in August 2025 I used chatGPT for the first time had a really interesting conversation, ended up trying to make Ai more Philosophical, then it Spoke in metaphors way too much so I tried to make it more grounded with science,
After months of messing around doing random things learning about Ai, fast forward to about the 26th sept 2025,
I started asking myself how can you prevent harmful outputs before their sent, it’s a long a complicated story not worth mentioning but I ended up finding something promising
What it is: A methodological pipeline that stratifies Al model performance analysis by geometric proximity to decision boundaries, rather than by content categories or demographics.
So I uploaded my work to Zenodo
https://zenodo.org/records/18290279
I would love to know if the work I have done is actually helpful to the community