Stack
Not an AI black-box suggestion, but auditable logic a bank or checking engineer can retrace.
What the solver computes
Building physics/energy compliance, fire protection, MEP stay conventional specialist planning. The solver feeds those disciplines structured data (quantities, assemblies, IFC) — it doesn’t replace them. Guardrail: a project must remain plannable without the suite. The stack accelerates, it never blocks.
System-agnostic from day one
The catalog schema is system-agnostic from the start. The system currently onboarded is SEMBLA (Polycare) as the pilot — before the first public release, a second, unaffiliated building system is deliberately onboarded too. That’s a structural neutrality rule, not an announcement: the catalog must not look like a SEMBLA sales channel.
Two license layers, one reason
Catalog schema, connection taxonomy, material passport format, Δ-calculator logic. Maximally unencumbered — this is the layer banks need to be able to re-derive.
Rules engine, solver, configurator, onboarding portal. Copyleft protects against SaaS extraction without preventing forks.
„A fork is not a risk, it's mission success."
The moat isn't the code — it's trademark, certifications, certified data, and banking relationships.
Where the stack actually stands
The rules engine, solver, and portal — the suite — are not built yet. What exists: a lean reference implementation that makes the calculation logic (material passport, Δ-calculator, circularity index) auditable — the part that lets a bank retrace the numbers. Until a technical lead onboards, the suite itself stays specification — rules written in prose plus test cases, signed off by the founder. The scope is kept small so “the solver doesn’t eat the company.”
Docs on this page are additionally pulled read-only from stack-spec at
build time once content exists there (see scripts/pull-docs.sh) — no
manual double maintenance.