Six companies. One operating thread.

Each company sat in a different domain, but each one forced the same underlying question in a different setting: how does a system stay coherent when real users, real failures, and real constraints push back on the design? Kwippy asked it of event streams and presence; Jaja.tv asked it of real-time, live-audience media; Octo.ai asked it of machine-learning workflows; ExpressMOJO asked it of fragmented physical-world logistics; Boom asked it of multi-chain infrastructure; Neul Labs asks it now of AI agents that can take real action.

Reading the six in sequence is less a chronology of separate companies than one long argument about where systems break — and it is the same argument that now runs through the governed-agent, durable-execution, and evaluation work at Neul Labs.

The domains changed every time — social presence, live broadcast media, machine-learning infrastructure, physical logistics, multi-chain finance, agentic AI — but the founding discipline stayed constant: understand the operating environment before designing the interface, treat failure modes as a first-class design input rather than an afterthought, and be explicit about which claims are supported by evidence and which are still aspiration. That discipline is why the individual company pages here are deliberately conservative about numbers that are not yet backed by a linkable primary source, even where an earlier version of the same story was told less carefully.