An estate that cannot describe itself
Most organisations still describe their own technology estate in a spreadsheet. It is wrong the moment it is saved, because the estate keeps changing and the document does not. When something breaks, the first hour is spent not fixing the problem but reconstructing what was even connected to what.
The July 2024 CrowdStrike outage made the cost of that gap concrete. A single faulty file reached every endpoint at once. The institutions that recovered in hours could see their inventory and their dependencies. The ones that took days were rebuilding both from memory while the clock ran. The difference was not luck. It was whether the estate could describe itself.
What if the estate drew itself
Now picture the opposite. You point a system at a network and walk away. Small agents fan out, resolve every host, every service and every link, and the picture assembles itself in front of you: live, in three dimensions, redrawing as each new node appears. No analyst transcribing boxes into a diagram that is stale before the meeting starts. The map is the territory, and it updates itself.
That is the demand on the table today. Live inventory, a current architecture graph, a model you can compute risk on. It reads like a roadmap item for the late 2020s.
This is not a concept render. It is 2011.
The work in this footage dates to 2011, with the video published in 2014. It is a distributed, parallel program written in Limbo, the native language of Inferno, the operating system from Vita Nuova. Identical agents run concurrently across the network, each performing discovery and inventory on its slice, and each streaming what it finds into UbiGraph, a 3D graph engine that redraws the topology in real time as hosts and links resolve. Nobody is drawing the picture. The network is drawing itself, and the graph is the output of the computation, not a slide made afterward.
Right idea, exotic tools
Two of those three components are now museum pieces. Limbo and Inferno never reached the mainstream. UbiGraph has been dead for years. It would be easy to file the clip under nostalgia.
The choice of tools was deliberate, not eccentric. Inferno treated distribution and concurrency as first-class properties rather than libraries bolted on later, which is exactly what an estate-wide discovery sweep needs: many small jobs running at once, their results merging into one coherent view. The technology was a means. The claim underneath it was the real artefact, and the claim was simple. An estate can discover and draw itself faster and more honestly than any human can document it, and once it does, that live picture becomes something you can reason and compute on.
From that bench to what we ship
Three things from the 2011 prototype survive, intact, in the current product line. The live discovery and inventory loop is now the foundation the rest stands on: nothing downstream is trusted unless the inventory underneath it is current. The self-drawing graph grew up into Diagrammar, which produces layered 3D architecture graphs straight from live inventory and discovery data, so the business, data, application and technology views are always the real estate rather than a year-old drawing. And the insight that a graph is not just a picture but a model you can run turned into DORA-MAST, which models operational resilience on that same graph, and cVaR, which prices what it costs when one node fails everywhere at once.
The exotic runtime is gone. The architecture of the idea is not. We kept the shape and changed the engine.
The new normal
The line from the clip to the platform is short and straight. What looked like a research curiosity in 2014 is now the unglamorous discipline that decides whether a bad Tuesday lasts hours or days. Inventory is computed, not transcribed. The graph is current because it is generated, not maintained. And because the graph is a model, the question after an incident stops being "what was connected to that?" and becomes "we already knew, here is the priced blast radius."
Discovery was never the hard part. Believing that a live inventory mattered enough to build the whole stack around it was. We placed that bet in 2011, on film. The rest of the industry is arriving now, fifteen years later.