ARTEMIS runs design, estimating, engineering and field delivery on one live 5D model — geometry, engineering calculations, material specifications, RFIs, correspondence and labor productivity in a single synchronized system, orchestrated by OctopusG and verified against real standards. Version A2 shows that stack in motion: what it is, how it is built, and the figures it is held to.
Parametric building geometry — duct, pipe and equipment as objects with properties, not lines on a sheet.
The known logos, arranged the way the work actually flows: OctopusG at the center directing the panel, 9Router in tight orbit as the routing hub, the AI engine models on the first ring, the repositories, agents and deployment that carry their work on the second, and the languages and libraries it is all built in on the outer ring. Select any mark to read its role — or switch to the index for the full tiered list.
RINGS, INSIDE OUT: OCTOPUSG — CLOSEST ORBIT · 9ROUTER · AI ENGINE MODELS · REPOSITORIES, DRIVES, AGENTS, SERVERS & DEPLOYMENT · PROGRAMMING LANGUAGES & LIBRARIES. CLAUDE AND ANTHROPIC ALSO SERVE AS THE AGENT & SKILL RUNTIME — SEE THE INDEX.
MARKS: SUPPLIED BY THE HOUSE (ANTHROPIC · CLAUDE · GOOGLE · DEEPSEEK · PYTHON · MAPGL · JAVASCRIPT) AND OPEN PUBLIC LOGO SETS (SIMPLE ICONS, CC0) FOR THE REST — REAL MARKS, LABELED, NOTHING REDRAWN. GROK, 9ROUTER AND KATEX CARRY TYPOGRAPHIC TILES WHERE NO CLEAN PUBLIC MARK EXISTS.
A 5D model is only useful if every discipline reads the same object. In the ARTEMIS platform, a duct run is simultaneously a drawn element, a calculated load path, a weighed quantity, a priced line, a scheduled activity and a field instruction. Change the geometry and quantities, labor, duration, cost, schedule of values and cashflow recalculate in the same pass — there is no export, no re-keying, and no second version of the truth to reconcile later.
| MEASURE | BASE | P80 (×1.16) |
|---|---|---|
| Sheet metal quantity | 186,000 LB | — |
| Cooling capacity | 850 tons | — |
| Direct cost | $14,143,700 | $16,406,692 |
| Schedule of values | $16,123,818 | $18,703,629 |
| Duration | 67 wd | — |
Snapshot scales linearly from the verified HVAC Bridge base case (186,000 LB at 120,000 SF). SOV = Direct × 1.14 in all cases. Full model: HVAC Mechanical Intelligence Bridge.
“The meeting starts at the decision, not at reconciling whose numbers are right.”
The stack is organized the way the work is organized. AI models reason, context & code build, the core orchestrates, design & CAD author the engineering, and delivery & commercial systems carry it into contract, field and payment. Each lane feeds the next; OctopusG directs all five.
Multiple frontier and open models on one routed panel — each task goes to the model best suited to it, and panel outputs are cross-checked before they are used.
The engineering toolchain: structured context, code, typeset mathematics and real-time 3D — the layer where models, calculators and simulations are actually built.
The center of the system: the orchestrator, the architecture it enforces, and the domain the whole environment is published under.
Where the engineering is authored: parametric and 3D modeling under disciplined CAD management, in the Autodesk environment the industry builds from.
Where the model meets the contract: enterprise construction management, field coordination, payment processing — and the business document store that feeds the live model.
Production hosting and the release pipeline. Every ARTEMIS application ships through Vercel's Git-connected deployment — reviewed, versioned, and rolled out the same way every time.
Identity and access for the team, and the search and discovery layer across the business's documents and data.
The document source of truth: drawings, specifications, estimates and correspondence live in Google Business Drive and feed the live 5D models directly.
ARTEMIS does not pass work down a line. The live 5D model, the AI engines and skills, OctopusG and its agents, the stack and systems, and the teams in New York, Cyprus and the field all work the same job at the same time. Scrub through the job below: at any instant, more than one phase is live — procurement starts while estimating is still sharpening, and closeout lessons feed the next bid while construction is still running.
Where the bars overlap, the phases are running concurrently on the same model — the next phase is already being built while the current one is still live.
Claude leads takeoff and estimate reasoning; DeepSeek stress-checks the cost build-up; Grok argues the no-go before the panel commits to a bid number.
Gemini reviews submittals and specifications against the model objects they approve; Claude drafts procurement scopes; the full panel runs the pre-buyout risk review.
Engines resolve RFIs on the objects they block; DeepSeek runs cost-to-complete analysis; Grok challenges every change-order price before it is submitted.
The panel audits final cost against the estimate, line by line, and extracts the lessons — what moved, by how much, and why — in writing, not from memory.
OctopusG opens the pursuit record the moment documents land; the Attention Engine flags deadlines and addenda; the Decision Center holds the go / no-go.
Kickoff is orchestrated, not emailed: owners, tasks and evidence states are assigned on the model; DEV agents build the job setup, JEV verifies it before mobilization.
The Attention Engine surfaces what changed today — a slipped delivery, a productivity drift, an unanswered RFI — and the Decision Center routes each item to a decision.
OctopusG compiles the review pack and promotes verified lessons into the shared context — so the next pursuit opens with this job's facts already loaded.
Drawings and specs land in Drive; takeoff runs on the 3D model in Three.js + Python; every estimate version is committed in GitHub.
CMiC carries the job cost breakdown; Procore runs submittals and buyout; the SOV is set for Textura billing before the first requisition exists.
CMiCPROCORETEXTURAProcore holds field, RFIs and punch; CMiC holds job cost; Textura moves progress payments; model views publish through Vercel for every stakeholder.
PROCORECMiCTEXTURAO&M manuals deliver from the same Drive the job ran on; retention releases in Textura; the final record archives in GitHub + CMiC, complete and auditable.
New York leads estimating and bid strategy; Cyprus checks engineering feasibility the same day — no overnight hand-off, the question is answered while the estimate is still open.
NEW YORK · ESTIMATINGCYPRUS · ENGINEERINGCyprus details coordination and calculations; New York locks the commercial terms; the field validates means and methods against real site conditions before buyout.
CYPRUS · ENGINEERINGNEW YORK · DESIGNFIELDThe field runs live production against the model; Cyprus turns engineering answers around in hours; New York holds cost, schedule and the client — all reading the same numbers.
FIELD · LIVE PRODUCTIONCYPRUS · ENGINEERINGNEW YORKThe field proves the systems and burns the punch list to zero; all three teams sign the lessons review together — the people who estimated it, engineered it and built it, in one room.
FIELDNEW YORKCYPRUSA model that cannot be checked against a standard is an opinion with good graphics. The ARTEMIS bridges read every design through published standards lenses — and the stack's own figures are held to the same discipline: reproduce the verified base case exactly, or declare the deviation.
The reference implementation of the platform: a live HVAC 5D model — geometry, quantities, schedule, field production, cost and executive views on one object-linked system, and the source of the verified base case on this page (RC8.11).
The same discipline applied to utility infrastructure: a live 3D utility model carrying quantities, trench and pipe engineering, schedule and cost — checked against the standards that govern work in the public right-of-way.
Following the OctopusG discipline, every fact the system presents carries its standing. A number marked Verified has been reproduced against its source; a number marked Declared is an intention or a statement, not a measurement. The distinction is never blurred — it is the difference between the five operating bases below and the 2027 sequence, and between a base case and a forecast.
The stack is not a headquarters tool that offices log into — it is the same live environment in every base, which is what makes a Cyprus engineering team and a New York estimating team one team. Five bases operate on it today.
Design + estimating headquarters. Client-facing decisions, executive review, commercial control.
Engineering center. Model authoring, calculations, CAD management and production engineering.
Regional engineering and design support on the same live models.
Regional base supporting engineering delivery across the Aegean program.
Regional base extending the network into Italy on the shared platform.
A stack of strong tools still needs a conductor. OctopusG is the top agent orchestrator of the ARTEMIS environment: it routes work across the AI model panel, holds the context every agent shares, and runs the Command → Universe → Attention → Decisions → Intelligence discipline — so AI output arrives as evidence and interpretation, never as unattributed authority, and every recommendation can be traced, challenged and verified.
Routes each task to the right model on the panel — Claude, Gemini Astra, DeepSeek, Copilot, open source — and reconciles their outputs against one shared context instead of letting five answers compete.
Builds and maintains the structured context — models, documents, decisions, evidence — that every agent and every human on the team works from. Same inputs, same conclusions.
Owns the architecture of the whole environment: how the lanes connect, where data lives, and how a change in one system propagates safely to the rest.
Deterministic and auditable by design: scans the universe of projects and surfaces what actually needs a human decision — ranked, evidenced, and quiet about the rest.
Every surfaced item carries an evidence state — Observed, Declared, Inferred, Verified, Blocked, Stale or Unknown — so executives decide on facts whose standing is explicit.
Coordinates multi-agent work and verifies it before integration: figures must reproduce exactly or be declared as deviations. Nothing merges on trust.