Back to homenorthROSC LABS · Deepchox

Product guide

How your AI-powered team for founders works

Built for founders: one venture, one surface, AI teammates at each desk that stay aligned with how you run the company.

The solo-founder loop

  1. Anchor a venture

    New ventures open in Personal Assistant: you describe the idea in chat; the assistant asks follow-ups (including tap-to-answer options) and writes into your venture record so every desk shares the same ground truth.

  2. Work desks, not tabs

    Move between CEO, product, finance, and scout as jobs appear—artifacts stay in one venture record.

  3. Let staff sync compress research

    When you run sync, the model stack updates market and desk notes so you are not copy-pasting between chats.

  4. Run Dexo for a full cross-functional read

    Open the Dexo room from the floating orb or nav. Tap Analyse — Dexo reads every venture field and speaks a structured briefing aloud. Interrupt it at any point to ask a follow-up; it picks up the thread without restarting the analysis.

  5. Read the office from the dashboard

    Execution signal, phases, and desk coverage give you a single place to see whether the plan is moving.

How agents behave

  • Desk-scoped: each role has a mandate; outputs read as memos and artifacts, not random Q&A.
  • Venture-consistent: strategy, backlog, and numbers reference the same venture so narrative does not drift.
  • Syncable: staff sync is the batch pass that folds fresh research into your operating picture.

Dexo Intelligence Room

  • Canvas sphere orb: a live 60 fps particle sphere renders at the centre of the room. While Dexo is speaking, the sphere pulses and shifts colour in sync with the voice — three overlapping sine-wave frequencies drive organic motion, not looping CSS keyframes.
  • Non-blocking analysis: clicking Analyse never locks the UI. The message bar and voice button are available immediately. Type or speak while the briefing compiles; the thread updates when results land.
  • Jarvis-style interruptible voice: Dexo reads its report aloud. Tap the mic or speak at any moment — the current utterance cancels, the interrupt is logged in the conversation, and your input is processed immediately.
  • Inline controls: Mute toggles voice without closing the report. Re-read replays the full briefing. Re-analyse reruns the full cross-functional pass against updated venture fields.

Floating Dexo orb

  • Persistent entry point: the animated orb floats at the bottom-right of every screen. Tap it to enter the Dexo room instantly — no nav required.
  • Voice greeting: on first tap, Dexo speaks an introduction. On subsequent taps it acknowledges your return. The orb remembers across sessions so it never repeats the full intro.
  • Reactive canvas: the orb idle-rotates slowly and breathes gently. On hover it speeds up with a cooler palette; thin motion rings show it is ready — no heavy glow.
  • Draggable: drag the orb to any corner — its position persists across sessions via local storage.

Claude × GPT — how models work together

Deepchox uses a dual-provider layout: GPT drives most desk conversations with role-tuned system prompts, while Claude runs in the same staff-sync pipeline so both families contribute to merged outputs. The Intelligence Suite shows a Claude × GPT panel while sync runs. Dexo adds a full-record analysis pass. You never choose a model in the UI — routing is automatic and follows your backend configuration.

FAQ

Modeling and behaviour — without deployment trivia.

Deepchox runs a Claude × GPT dual stack. OpenAI GPT powers desk agents (CEO, CFO, PM, CMO, Scout) with role-specific prompts. Anthropic Claude runs in parallel on staff sync alongside GPT so both model families contribute to merged desk briefs. The Dexo Intelligence Room adds a dedicated analysis pass across your venture fields. Routing is automatic — you never pick a provider.