Dev Release #9 — The platform gets a mind, and a third pillar

The platform gets a mind, and a third pillar.

Four moves this release. The financier seat went from design to working software, and Trade OS gained the layer that changes what it is: Atlas, the Brain, and Professional Services AI — the system of record became a system that works for you.

Growyn grew from an introduction into a running engine, section by section, including a video studio that writes, voices, and cuts. The rail has its audits signed, its purchase flow on Ethereum mainnet, and its testnet date locked. And we are introducing energyOS — the third pillar.

Part 1 — Trade OS: the platform, portal by portal

One platform, one record of the trade, and a window into it for every party. Here is where each portal stands, how they work together, what the AI layer does, and how the documents engine runs.

One trade, every seat at the table. — Platform: Complete

Trade OS holds the entire trade lifecycle as one connected record — products, manufacturers, clients, orders, production, shipments, finance, documents — with an event-driven double-entry ledger wired underneath, so landed cost and margin are live per order, not reconstructed at quarter end. Every counterparty portal is a scoped window into that same record: nobody gets a copy, everybody gets their view.

The Operator portal. — Complete

The desk that runs the trade: orders with a two-field status model, production tracked in stages per product family with multi-variant lots, shipments with carrier tracking, finance from invoice to payment to cashflow and letters of credit, a task board, pipeline, and messaging threaded to the entities they are about. This is the portal EDMA Group runs its own operation on.

The Supplier portal. — Complete

The manufacturer's window: purchase orders to confirm or counter, production milestones to report, a document-requests inbox where the operator's requirements arrive and are fulfilled in place, and proforma flows through the same document engine.

The Client portal. — Complete

The buyer's window: their orders with live status, shipment tracking, documents scoped to what they are allowed to see, invoices, and a claims path — raised, investigated, resolved on the record, not in an inbox.

The Logistics portal. — Complete

The forwarder's window: shipments to advance, documents to file against them, container, bill-of-lading and shipment-number scan lookup, and driver mobile flows — a stops list with status progression and proof-of-delivery capture in the field.

The Financier portal. — Complete — new this release

In #8 this was fifteen screens of design; it is now working software, and the fifth seat is occupied. A financier registers and passes vetting self-serve — sanctions screening included — then works a desk built around disclosure as architecture: published receivables appear summary-only by default (industry, regions, value band, terms, and the operator's reputation band, recalculated nightly); full detail opens only after an explicit disclosure handshake.

From there: compose an offer, track it to a decision, and manage the agreement from pending-legal through active to settled. Deals sourced outside the platform enter through a guided wizard and land in the same book, so the portfolio view — capital deployed, yield, repayment behavior, concentration — covers everything.

How the portals work together.

One order ripples outward: the operator confirms it, the supplier sees a purchase order, production milestones flow back, the forwarder's shipment updates the client's tracking, the invoice reaches the client, and — if financed — the financier's agreement settles against the same record.

Counterparties join by magic-link invitation with entity-scoped permissions; every portal sees the same truth, redacted to its role. Financing coordination runs across three seats: the operator publishes, the supplier sees its receivable position, the financier funds.

The Documents engine — one engine, every portal. — Complete

Trade runs on paper, so the paper became one system. Upload anything and intake intelligence reads it — OCR, extracted fields you accept or correct, full-text search across the archive. Documents generate themselves from live order and shipment data.

A real editor sits on top: e-signatures with saved libraries and certified copies, version history and compare, comments, watermarks, audit trails, and secure expiring links for anything that leaves the platform. Operator, supplier, client, logistics, and financier each render the same document through their own scoped route — one source, five lawful views.

The AI layer: Atlas, the Brain, and Professional Services. — Complete

Atlas is the platform's agent — ambient across every portal with a command bar, a needs-you queue, and a handled feed. It watches the business on schedule — stalled orders, overdue shipments and invoices, expiring certificates, stock levels, margins, client health — and raises signals before you go looking. It acts through gated tools: propose, approve, done — approval authorizes the action itself. Every specialist sits on a trust ladder you set per capability; Watching structurally blocks action. It drafts in the counterparty's language, and a coverage page says plainly what it can and cannot do.

The Brain grounds it: a prediction ledger with a visible track record — including the misses — calibrated facts learned from your own operation, and trusted context injected into every AI answer so nothing is invented.

Professional Services AI does the work you would send out: contract analysis with confidence scored by the platform rather than the model's self-report, contract drafting with history, and a clause library built from your own agreements; accounting categorization — single or batch — grounded in your chart of accounts; and P&L, balance-sheet, and cash-flow commentary on the live ledger. The full three-statement set, exportable.

On the dashboard: a Daily Brief assembled from open signals, a predictions panel that shows its own track record, and Ask — plain-language questions over your operation, answered by a structured query engine across your entities, never a made-up number.

Automation with a safety system. — Complete

Bot Studio in guided and manual modes, a library of twenty outcome-named helpers you add in one click, and dry-run previews that show what a helper would do before it touches anything. Underneath: retries with backoff, error-rate auto-pause, priority rules between conflicting workflows, rate limits on bulk triggers, and circuit breakers on workflow cascades.

The platform runs as a business. — Complete

An operations portal for the platform itself: tenant management with suspension enforced across every portal, impersonation with full transparency and instant revocation, versioned pricing with subscriptions, invoicing, refunds, credits and dunning, GDPR tooling and legal holds, and a full audit surface — with two-admin release and re-authentication on sensitive actions. Alongside it, a security hardening pass across the platform: input validation on every finance route, global rate limiting, and credentials encrypted at rest.

edma.trade. — Complete — whitelist open

The product's home now shows the product working: self-playing simulations of the supplier, logistics, client, and financier portals on the homepage, and a listing moving from accept-best-offer to funded — in seven languages. Visit edma.trade →

Part 2 — Growyn: the engine, section by section

In #8, Growyn was an introduction. Since then the suite went from designed screens to working sections across the whole loop. Here is the engine, section by section.

Pre-launch · doors-open track

Intelligence — the research department. — Complete

The starting point: a multi-step research pipeline with live web search that produces the full strategic picture — market, audience, competition, demand, pain points, pricing, brand, channels, SWOT, tradeoffs, validation, synthesis — as living sections rather than a PDF dump, with a confidence layer that says how sure each claim is, and a whitepaper generator for the businesses that need one. Industry packs adapt the entire run to the business detected — a blockchain launch and a local business get different channels, deliverables, and compliance out of the same pipeline. Re-runs version cleanly.

Funnel Plan — the marketing brain. — Complete

Where the whole plan lives: a milestone-gated hub, a Funnel Builder that models stages, channels, conversion estimates and budget — with real ARPU sourced from actual pricing, transparent reach sources, and what-if scenarios — and a long-form exportable Full Plan. The plan is not decoration: it issues directives the other sections consume, and their results report back against it, real-versus-expected per stage. Locked plans version.

Website — generate it, or bring your own. — Complete

Two paths. Growyn generates a site — five-stage AI generation, block editor, blog, theming from the brand kit, SEO and tracking wired in, publish to hosting. Or point it at the site you already have: the imported-website editor connects to a GitHub repository and you edit visually — the agent sees exactly what you selected, small visual edits apply instantly through a deterministic parser, bigger ones route through the model, and everything ships as a pull request you approve. A second, bespoke generation engine produces hand-written-grade multi-page sites with populated product mocks and engineered diagrams.

Content — the newsroom. — Complete — publishing connectors rolling out

The week plans itself: "Generate my week" decides channels, pillars and cadence from the plan, writes everything, scores every piece, and runs a deterministic number-audit so no unverified statistic ships. Around it: a calendar with month and week views, a library that holds every asset including video, a review queue, and revenue attribution down to the piece through the shortlink engine.

The Creative Studio. — Complete

One conversational pattern across writing, images, and video: a Creative Session where the work is debated before money is spent — concept before spend — grounded in what the Brain knows about the business, with every paid generation kept as a variant you can return to.

The video studio. — Complete

A full production pipeline on Growyn's own stack. Storyboards written under explicit story-logic laws — one meaning per symbol, lines true to their frames, briefs limited to what a camera photographs in one frame. A professional full-page editor with scene lanes, per-scene directed voiceover, and punch captions timed word-by-word to speech. Characters speak with lip-sync on our own model queue — no avatar SaaS. A talking-head lane for real founder footage. Sound effects and a built-in music library. And film grammar in the render: sound bridges, cuts on motion, dissolves reserved for time jumps.

SERIES. — Complete — new

Episodic content as a product: each customer runs a weekly stylized series starring their recurring cast, written by the engine from their real audience research. A series bible, a writers' room that generates whole seasons behind an approval gate, an episode engine, fixed voices per character with lip-synced dialogue, photo-form episodes — carousels, comics, banded slides — covers baked as frame one, teaser, quote and poll derivatives mined between drops, and per-series measurement. Story-shaped, not sales-shaped: the product is a prop in the world; the season arc carries the transformation.

SEO — an engine, not an audit tab. — Complete

A continuously updated keyword map fed by Intelligence demand signals and enriched with volume, difficulty, and trend; Search Console tells it what you already rank for and what is moving. A research-grounded article writer produces evidence-cited pieces that link only to pages that actually exist, and an autopilot plans and writes on a cadence you set — draft or auto-publish, to your Growyn site or WordPress, Webflow, Shopify, Ghost, Wix. Site audit, backlinks, and visibility tracking round it out.

Paid Ads — the budget optimizer. — Complete

A creation flow that takes a campaign from type to format to creative to launch, across mainstream and crypto-native channels — Google, Meta, LinkedIn, TikTok, Reddit, YouTube, X, plus CoinTraffic, CoinZilla, BitMedia, and Brave. The overview is a budget optimizer with three honest states: cold — no data yet, so it shows the plan, not results; warm — campaigns live but under seven days, so real numbers, no advice yet; live — recommendations and money moves. Scale, reduce, hold and pause calls are a deterministic rules pass over CPA, ROAS, target, and trend; every estimate carries its provenance. Boost turns winning organic content into spend.

Capture — traffic into contacts. — Complete

Four asset types — forms, popups, lead magnets, surveys — plus polls, meeting booking with routing rules, and progressive profiling for return visitors. The visitor side is a pure-code pipeline: rate-limited, email-validated, disposable-domain-blocked submissions fan out to the contact database, attribution, and nurture enrollment in one motion. The creation side drafts with AI grounded in the Brain, and the autopilot fills the gaps the funnel plan identifies — in Automated mode it publishes with undo.

Nurture — journeys that run on clockwork. — Complete

An email system with a visual journey builder where the clock and the branching are pure logic — AI writes the emails; it never runs a journey. Contacts enter by form, event, segment, or hand; re-entry is governed by cooldowns and caps; send policies enforce daily caps and quiet hours; suppressions and consent gate every send through an idempotent ledger. Segments, scoring rules, templates, sending domains, imports, and reports complete it.

Engage — every conversation in one inbox. — Complete — channels expanding

One channel-agnostic conversation model behind every surface. The website chat widget runs the full loop — visitor message, AI auto-reply, human handoff with suggested responses. Bot Studio builds the personas, with a test sandbox. Community management runs on Telegram with rules-based moderation in review-only mode; reviews are captured with AI-drafted responses; insights show volume, sentiment, response coverage, and community health.

Close — the CRM. — Complete

Nine surfaces on a shared engine: pipelines with drag-and-drop kanban and rot and hot flags, companies and contacts with 360 views and account heatmaps, prospecting as a live work queue, campaigns with A/B testing and precise per-campaign attribution, activities fed by sequence-created tasks, sales sequences on the same consent-gated engine as Nurture, deterministic lead scoring, revenue KPIs computed server-side — nothing AI-invented — and a workspace with a live brief, next-best-actions, and Ask.

Analyze — the truth layer. — Complete — connecting sources

Overview, channels, funnel, visitors, reports, and a translator that turns analytics into plain language. The mode model is honest by construction: empty, simulated, live, and projected states are labeled as exactly that; ROAS shows a dash rather than a fabricated zero when no ad platform is connected; no intermediate funnel stage ever gets an invented actual. Reports generate against an honesty contract and share through revocable public links.

Optimize — the weekly discipline. — Complete

Scheduled workers detect patterns, propose actions with expected impact, measure outcomes after a window, and write the Monday brief. Auto-apply is governed by the operating mode, and every action lands in a ledger with its result — so the system's advice has a track record you can check.

Assets and the ContentDNA library. — Complete

Brand kit — logo, fonts, colors, kit health — media with auto-tagging, templates, and exports. And ContentDNA: the reusable hooks, frameworks, CTAs, and topics Growyn extracts from analyzed content and feeds back into everything it generates.

The platform layer: one Brain, three modes, one inbox. — Complete

Every AI generation in every section routes through one shared Brain — a single context-assembly entry point that injects what the whole system has learned into each task, with per-run cost accounting on every piece of work the platform produces. Operating modes — Manual, AI-Assisted, Automated — are a platform primitive: each section can propose, draft for approval, or act with undo. Everything pending lands in one approvals inbox. First-party tracking and conversion definitions run through the Spine; the data hub syncs connectors on schedule; billing meters credits against real usage.

growyn.co. — Complete

The public site is live.

Part 3 — The rail: audited, dated, and taking purchases on mainnet

Testnet opens August 20 · mainnet Q4 2026

The contract suite — PoV Gate, One-Claim Ledger, attestor registry, EDSD, EMT, settlement controller, fee router, burner — stands deployed on testnet with both security audits signed. The $EDM purchase flow now runs on Ethereum mainnet with live pricing. Public testnet opens August 20 with attestor onboarding; mainnet promotion follows in Q4. The sequencing principle holds: Trade OS proves the OS, the OS proves the rail, the rail proves the marketplaces.

Part 4 — Introducing energyOS

Blueprint complete · cockpit designed · Phase 1 starting

The third pillar, alongside Trade OS and Growyn: an AI operating system for energy assets. One immutable ledger takes in everything a park produces — telemetry, market prices, bank feeds, contracts, permits — and every participant looks at that same ledger through a role-filtered window. Every number traces to its source artifact in one tap: a meter reading, a bank statement line, a signed invoice. That is what turns live data into data a lender will rely on.

Five AI bots run the park under a graded autonomy model: CFO — the project finance model as a live engine, covenants monitored continuously, lender packs generated rather than assembled; Ops — production against P50, loss attribution; Asset — predictive maintenance, battery health; Trading — dispatch and certificate monetization; and Compliance/ESG. Seven portals sit on the ledger — owner cockpit, lender, investor, EPC, offtaker, advisor, and a cross-project portfolio view for banks.

The owner cockpit is designed — home, plant, finance, maintenance, construction, timeline, tasks, dataroom — and the finance view already answers the questions that matter on one screen: is covenant headroom shrinking, is there cash to distribute, are the reserves funded. Customer-zero is our own anchor asset: a 237 MWp solar-plus-storage project whose finance model becomes the system's schema. Phase 1 puts the CFO Bot on that model, with the cockpit and the lender and investor portals — the same customer-zero discipline that built Trade OS.

What this adds up to.

Eight releases built the trade. This one gave the platform a mind — an agent that watches, proposes, and acts under your control; professional-grade legal, accounting, and reporting on your own data; a marketing engine that produces finished work, section by section; and a third pillar entering the build with a lender-grade ledger at its core. The rail underneath is audited and dated.

Verify first. Then mint.