Glen Healy

The throughline in Glen’s career is closing the gap between a person and what they’re trying to do.

He started in the US Army in 1999. First as a military police investigator in Hawaii, then as a recruiter in New Jersey, where he finished top 10% worldwide and earned the US Army Recruiter Ring. Eight years in, he got out, went to Rider University on the GI Bill, and graduated cum laude in multimedia communications.

Ten years of enterprise B2B sales followed. Verizon Business, then Comcast Business where he took over a Nashville team and built it into a top-five performer nationwide, then ADP.

Then COVID. Then his wife died giving birth to their fourth son. And the life that made sense before didn’t make sense after.

Glen stepped away from sales and took a job teaching sixth-grade English. Not as a pivot. As a reset.He needed to be closer to his four young boys in the shadow of their mom’s absence. Something closer to the ground.

His students had spent formative years behind screens, and it showed in their writing. Not in their ideas, which were as wild as any sixth grader’s ideas have ever been. In their ability to get those ideas onto a page in a way another person could feel. The gap wasn’t imagination. It was articulation.

One afternoon, on a hunch, Glen pulled up Midjourney and started generating AI images from his students’ sentences and showing them back to the class: this is what your writing looks like to someone else. The room changed. Kids who had been stuck started revising. They could see the distance between what they meant and what they’d written, and they closed it themselves.

That was the moment. Not an edtech revelation. Something smaller and more honest. A tool that could let people build what, until then, only existed inside their heads. That afternoon became the seed for Write With Riley, still on the Nuclear Marmalade roadmap. More importantly, it was the moment Glen committed to AI as the rest of his career.

He returned to telecom sales to pay the bills. Frontier, then Spectrum. But the real work was happening at night. The first problem he hit was the one everyone hits: ChatGPT has no memory. Every conversation starts over.

So Glen went deep. He taught himself vector databases, embedding models, retrieval architectures, token-budget ranking, memory-type weighting. The first system was naive. The second was better. The eighth became Nukaplakia.

He set out to build a second system he’d been sketching on paper for weeks: an inference layer that noticed when unrelated signals from different domains started clustering around the same thing. He called it Convergence Detection. Six engineering phases, shipped in a single week in March 2026. A sharp analyst might notice this kind of signal by accident after a decade on the job. Nuke does it on a schedule.

In March 2026, Glen left Spectrum and committed to Nuclear Marmalade full-time.