We built ARGUS to test a problem we solve in our everyday work: stitching together siloed tools into one system that works. Here, we applied it to a more technically demanding system.
Deciding where and how to build an outpost usually involves four specialists using four different programs: a terrain analyst maps the ground, an intelligence officer maps likely approach routes, a strike planner works from printed tables, and an engineer works in design and cost software. Each does their job well, but in isolation.
That gap shows up the moment a question touches more than one specialty. Raise the watchtower three meters and visibility improves, but nobody in the room can say how much the construction cost rises, how visible the post becomes to the other side, or how many minutes earlier an approaching group would be spotted.
The four specialties don't just use different software. They update on different schedules. A post's line of sight only changes when the structure itself moves, so it's slow and expensive to compute. Detection probability changes constantly, with every shift in weather, so it's fast and cheap. Combining both into one live system means solving each on its own timeline, not recomputing the slow part every time the fast part changes.
We combined four existing, working programs into one system, leaving each unchanged: a terrain and rehearsal tool, a tactical decision tool carrying military doctrine data, a site design and cost-estimating tool, and a reporting tool. Our roughly 22,500 lines of code are the layers built on top of them.
ARGUS is organized into eight different tabs, all reading off the same terrain and weather model:
Two principles anchor the system. First, the numbers always come from the calculation engine, never from the AI assistant used for natural language input. We explicitly barred the assistant from inventing a coordinate or a placement. Second, every fix goes all the way through: when ARGUS flags a surveillance gap, it also computes a recommended location to close it.
Across all eight tabs, three results stand out: Placement, Surveillance, and Operations analysis.
In one test run, the placement engine designed four structures and laid out the full site from scratch in under three seconds, raising joint sensor coverage from 0 to 49%, at an estimated cost of $1.7 million.
It also flagged an approach route with a complete surveillance gap, then computed a specific new observation position that raised coverage on that route from 12% to 54%, while staying outside enemy range.
During weather simulations, ARGUS recommends sensor swaps when visibility drops. On a clear day with radar on, every approach route is detected instantly. In night fog with radar off, detection slows by two to seven minutes and nearly half the coverage picture goes dark. At that point, ARGUS recommends switching to an acoustic sensor.
The terrain, sightline, and sensor modeling are real, and the rest of the system depends on them. The construction side is more early-stage: it currently uses a civilian building-cost model, so soil-hauling, drainage, and access roads are not yet modeled. Structural hardening is represented by a single placeholder number, not broken down into standoff distance or material grade. Lastly, the placement engine doesn't yet optimize for how visible a post would be to the other side; it tends toward good ridgelines because of how the terrain math works out, not because it's told to avoid exposure.
Ocean's Halo ran its packaging supply chain across spreadsheets owned by three companies, retyped by hand at every handoff. One dashboard now shows what needs attention, and the plants and the supplier enter their own numbers.
A newsroom's international desk used to lose hours every morning figuring out what was new. Now a tool does that overnight, and flags which stories the desk is likely to run.
A promotional-products franchise used to pay for a costly pricing tool with restricted control. Rian rebuilt it from scratch, handed over every line of source code, and made it four times faster in the process.
