Each product is not built from scratch. It's a thin domain layer on top of the Elastic Platform — the same compiler, database, protocol, GUI, and AI fleet that took 14 years to build. The remaining 10% is the product-specific logic that makes each one unique.
Every product inherits the full platform stack. The product-specific code is a DSM plugin — typically 5,000–15,000 lines of EPL that implements the unique business logic. Everything else is inherited.
Orion DB partitions = tables. TT_Partition + DataColumns + DataItems = the entire table system. Views, filters, formulas — all operate on existing partition infrastructure. User permissions, org isolation, real-time sync — inherited from DSM.
Column type system (30+ types: text, number, date, user, reference, formula, subtable). View rendering (list, board, calendar, Gantt). Data provider layer for dynamic schemas. Import/export (CSV). Marketplace app templates.
TProtocol handles all network I/O. Orion DB stores messages, folders, contacts, calendar events as partitions — same tree, same transactions. AI fleet classifies spam, auto-folders, executes actions. HSRV serves webmail. User auth, sessions, permissions — all inherited.
SMTP inbound/outbound (RFC 5321). IMAP server (RFC 3501) with IDLE. MIME parsing. DKIM/SPF/DMARC. iMIP calendar invites. Plus-addressing. TMailPlugin DSM plugin (~15K lines).
Orion DB stores repos, branches, commits as tree nodes — file system capability used directly. TProtocol handles git smart HTTP transport. AI fleet does code review via MCP tools. User permissions map to git branch protection. Org isolation = repository ownership.
Git pack protocol (upload-pack, receive-pack). Merge request system. Branch protection rules. Diff rendering. TGitHostPlugin DSM plugin. OrionGit streaming clone engine.
Orion DB partitions = shopping list, diet log, med schedule, budget, etc. — 17 partitions per user, all standard DSM partition/subtable infrastructure. AI fleet processes voice commands. Voice pipeline (Whisper + Piper) is a platform service. Mobile CEF wrapper is shared.
Voice intent router (deterministic regex + AI fallback). Dashboard UI. Open Food Facts / USDA nutrition database integration. Proactive rules engine (AIActionRules). On-device Android STT/TTS. Wake word integration.
AI fleet workers ARE Spark — they read the user's description, use MCP tools to create Tables partitions, define columns, insert data. The entire "app builder" is the AI fleet operating on the existing DSM infrastructure. Every generated app is just partitions + permissions.
System prompt engineering. MCP tool registration for app creation verbs. App template marketplace. Build confirmation flow. Per-app billing.
EPL compiler with spatial math (vectors, matrices, ray-surface intersections — added to the language). TProtocol for distributed network rendering. GUI framework for the 3D viewport. The entire 120K→15K reduction comes from platform capabilities.
Ray tracing core (BVH, photon mapping, adaptive sampling). Material/shader operator system. Scene graph. Camera system. ~15K lines of EPL. Read the full case study →
Kernel: Orion DB file system capability — files ARE tree nodes. Versioning = partition history.
Added: UI for browsing, upload/download, sharing links, preview.
Kernel: WebSocket push = real-time messages. Orion DB partitions = channels. User system = members.
Added: Chat UI, threading, typing indicators, channel management.
Kernel: V_Pipe video pipeline library (native C++, loaded by EPL). TProtocol streams video.
Added: Camera management, AI motion/person detection, alert system, recording.
A traditional software company builds each product from scratch — separate team, separate stack, separate infrastructure. Each product costs $500K–$2M and 6–12 months.
We build each product as a DSM plugin — 5,000–15,000 lines of EPL that inherits the entire platform. The 10th product costs roughly the same as the 2nd. The marginal cost doesn't grow with the product count — it shrinks, because each new product exposes platform capabilities that benefit all others.
When we added spatial math for Aurora, every product got access to vector operations. When we built the AI fleet for Spark, every product got AI capabilities. When we built the mail server, every product got email sending. The platform compounds. The products multiply.