|

The founding story

[This section to be written by Ari Aviv personally. Below is structural scaffolding — replace the bracketed passages with your own voice.]

[Where you were before Nexma. What you were building, what you saw that others missed. The specific moment — a project, a conversation, a failure — that made the problem impossible to ignore.]

[The early conviction: that AI, optimization, and spatial interfaces had matured enough to converge into a single platform. That infrastructure operators deserved better than toggling between CAD, spreadsheets, and disconnected GIS. That someone needed to build the operating system, not another point solution.]

[What you gave up to start. What the first weeks looked like. The decision to build in Tel Aviv.]

Influences

Three intellectual traditions shaped how we think about the problem. They are not trends we followed. They are convictions we hold, tested against the work we have done and the failures we have studied in the industry around us.

Spatial computing as a discipline, not a feature. GIS has existed for decades, but the industry has treated it as a visualization layer — something one looks at after the decisions have already been made. We reject that framing entirely. At Nexma, we treat spatial reasoning as the primary interface for design, optimization, and operations. The map is not the output. It is the workspace. This distinction is not semantic. It determines what the platform can do. A visualization layer displays results. A workspace produces them.

AI agents, not AI assistants. The shift from language models that answer questions to agents that take actions is the most consequential change in software since the internet. We built Jax as an agent from day one — not a chatbot with a map attached, but an autonomous system that reads, writes, and operates on the same data layer humans use. Conventional approaches bolt conversational interfaces onto existing tools and call it transformation. What Nexma has built is structurally different: the agent and the platform share a single source of truth, and the agent acts on it with the same authority as a human operator.

Skill-driven architecture. The insight that infrastructure domains share deep structural patterns — nodes, edges, constraints, optimization objectives — even when they share nothing on the surface. A fiber network and a water distribution system look nothing alike to their respective operators. Underneath, they are both constrained graphs with spatial embedding. One type system serves both. This is the architectural insight that makes the Nexma platform possible: domain specificity is a configuration, not a codebase.

Timeline

2025 — Nexma was founded in Tel Aviv. Development began on the skill-driven spatial platform, with the initial focus on proving that a single architecture could serve multiple infrastructure domains without code changes.

2025 — The core architecture was established: a persistent unified data layer, a skill-driven domain configuration system, and an autonomous AI agent. These three components remain the foundation of everything the platform does today.

2026 — The platform reached production readiness. The first vertical was completed as proof of architecture — demonstrating that a skill-driven agent could autonomously design infrastructure networks that met real engineering standards.

2026 — Marc Halbfinger, former CEO of PCCW Global, joined as advisor. His operational experience in global telecommunications infrastructure brought a level of domain scrutiny that strengthened both the product and our go-to-market approach.

2026 — The Book of Nexma was published. Ten enterprise products launched from a single codebase, each differentiated only by its agent skill.

What triggered building Nexma

[To be written by Ari. The structural argument below can stay as context, but the trigger should be personal.]

The trigger was not one event. It was the accumulation of evidence that three previously separate capabilities — large language models that can reason about engineering constraints, mathematical solvers fast enough for interactive use, and map rendering engines powerful enough to serve as design surfaces — had all crossed their usefulness thresholds within the same two-year window. The industry had not noticed because it was looking at each capability in isolation. We were looking at the intersection.

Any one of them alone produces a tool. All three together produce an operating system.

The question was not whether someone would build this. The question was whether it would be built by people who understand infrastructure — who have stood in a fiber trench, who know what a cascade depth violation means, who have watched a field crew work from a PDF that was already outdated when it was printed. Legacy tools are built by software companies that treat infrastructure as an abstract market. What Nexma has built comes from a different place: the direct experience of how spatial work is actually done, and the refusal to accept that it must always be done that way.

We decided it would be us.