This week’s biggest milestone was presenting the Black Knowledge Erasure Dataset (BKED) at Open Data Week, and it’s honestly still sitting with me in the best way.

I was part of the CUNY x Open Data Week takeover at the NYC PIT Pop Up at the Oculus. Presenting alongside MindHeart AI and SPARK**CIVIC made for an interesting lineup, and the drop-in format meant we were talking to a genuinely mixed crowd: researchers, practitioners, curious passersby who had no idea what they were walking into.

What struck me most was how immediately the project clicked for most people. I expected to spend a lot of time explaining what AI hallucinations even are, but most people were already primed to be skeptical of these systems. What caught their attention was the framing — treating hallucinations not as glitches to be patched, but as data worth studying, as a window into whose histories these models were trained to center or ignore.

The feedback and questions I got were genuinely energizing. People were impressed by the project’s scope, curious about the methodology, and a few had insights that are actually going to stick with me as we move into the data collection phase of our new dataset. There’s something clarifying about explaining your work to people outside your immediate academic circle; it forces you to articulate the “so what” in a way that a classroom setting sometimes doesn’t.

It has me more excited about where this project is going than I’ve felt in a while. Getting that kind of external validation early is a good sign. Now we just have to make sure the finished product lives up to the pitch.