Archerion Labs · independent AI lab
The AI is replaceable. The memory isn't.
I build memory for AI work, so the conversation you had yesterday still counts tomorrow. I'm a firefighter; my tools get held to that standard.
In the fire
2026ACTIVENASCMemory that outlives the model. Every session lands in one vendor-neutral workspace, so the next conversation, whether Claude, GPT, or something local, starts where the last one ended instead of cold. I run it on my own work every day; productizing it now.Read what's on the bench →Shipped
2026SHIPPEDSling SourceA medic scans a QR code at the scene and the patient lands on one page of nearby help: medical, senior services, crisis lines, each one tap to call or map. Built to stay legible and fast for a scared, elderly patient on a bad connection.Open slingsource.net →2026SHIPPEDSpeed ReaderPaste a wall of text and read it two to three times faster: one word at a time, with the pace easing off on the dense parts. Lives in the browser, and any AI that speaks MCP can hand it text directly.Open speedread →Who runs this
I'm RJ Sapp. I fight fires in Pasco County, Florida.
Before this, I built an app for the back of an ambulance. Something a medic could hand to a frightened, elderly patient at the worst moment of their day and have it just work: big text, no login, fast on an old phone with one bar of signal, still useful with the network gone. There was no version of "it crashed" that was acceptable. That standard never left me.
Now I run Archerion Labs at night: an independent lab building memory infrastructure for AI. The tools I ship are the ones I use on shift the next day. I'd rather ship one thing that holds than ten that don't.