Private beta · certification-gated · shadow-graded

Your autonomous
trading desk
that never sleeps.

Your wallet, your keys. Follow certified leads through one execution spine — every certification gate’s number is visible, every refusal is explained. An AI Risk Officer reviews every trade shadow-first: it can veto an order, never place one. Hard limits stop the bleeding if anything breaks.

1
execution spine
24/7
autonomous
Always Learning
The desk · certification-gated · autonomous loop
illustrative
Fundr
funding arb
standby
Momo
momentum rot.
standby
Vexa
convex plays
standby
Shadow
smart-money
standby
Echo
news sentinel
standby
Warden
risk warden
standby
THE DESK
Scanning.
one execution spine
Autonomous · Non-custodial · Always learning
Non-custodial
you hold the keys
1
execution spine, every gate visible
24/7
autonomous
Connecting to the desk…
status · live from operator api
engines running
last scan
hl latency
The shared learning loop

Every user makes the system smarter.

Every AI decision QuantDesk makes — on your book and everyone else’s — is graded after the fact. Decision types that keep failing across users are retired. Your positions stay private; the graded outcome is shared — anonymous, tenant-stripped, aggregated — so every user’s risk officer sees it.

The more people use QuantDesk, the sharper tomorrow’s signals get.

$ pattern libraryillustrative
funding>15% + momentum_confirm → short setup
single_source_news + no_confirm → avoid
lead_certified + controls_passed → copy
illustrative — how graded decisions are labelled
What’s inside

A trading desk that runs itself. Certification-gated. 24/7. While you do something else.

The desk works while you sleep.

One execution spine follows every certified lead in parallel — the certification gate's numbers (win rate, sample size, verdict) are visible before you ever follow one. An AI Risk Officer reviews every trade shadow-first: it can veto an order, never place one.

  • ·24/7 · Hyperliquid perps
  • ·Per-trade risk, daily loss, max drawdown — all enforced in code
  • ·Telegram pings only on real events, no periodic noise
You stay in control.

Non-custodial by design. Your main key never leaves your wallet. QuantDesk gets a trading agent — a sub-key that can place orders but cannot withdraw — that you revoke with one click.

  • ·WalletConnect one-click authorization
  • ·Revoke anytime from the venue’s official UI
  • ·Chat co-pilot speaks quant — “halt momentum, it’s chopping” just works
Everyone’s trades make it smarter.

The learning loop grades every AI Risk Officer decision after the fact and retires decision types that keep failing. Contributions are anonymous and tenant-stripped — your book stays private.

  • ·Private: positions, P&L, wallet, chat
  • ·Shared: anonymized outcome of every AI judgment
  • ·More users → sharper signals
How it works

Three steps to a live trading desk.

01
Connect your wallet

Paste your Hyperliquid wallet address. We verify read-only, then you authorize a trading agent key in Hyperliquid's official UI — revocable in one click.

02
Pick a strategy

Conservative or Balanced — or tune every dial yourself. Change it anytime via the chat co-pilot.

03
Go live

The desk goes live: every trade clears the certification gate before it fires. Telegram alerts every entry, exit, and kill-switch. You can halt or pause from anywhere, anytime.

Learn together · trade apart

What leaves your tenant — and what doesn’t.

your tenant · private
Stays inside your account
  • Wallet address · agent key (encrypted)
  • Positions · entries · exits · P&L
  • Equity curve · capital size
  • Risk dials · strategy presets
  • Chat co-pilot history
outcomes only · tenant-stripped
shared pool · anonymized
Feeds back into every user’s AI
  • ·Pattern titles + trigger conditions
  • ·Win-rate · sample size · time window
  • ·Graded outcome of each AI judgment
  • ·NO identity. NO wallet. NO P&L.
Read this first

The three things every beta user should know.

Can the system lose all my money?

Yes. Trading is inherently risky. QuantDesk copies certified lead wallets with your capital — a lead that passed certification can still lose. Hard kill switches contain runaway bugs — they don’t prevent market losses. Only use capital you can afford to lose.

Do you hold my funds?

No. Never. You authorise a trading agent (a sub-key) that QuantDesk can trade with — but cannot withdraw. You revoke it from Hyperliquid’s own UI at any time, no approval required.

What about bugs?

There will be bugs. This is beta. Kill switches and hard caps contain blast radius, and every AI decision runs shadow-first — graded before it is trusted. The risk is yours.

on waitlist·invited this week

Private beta. Invite only.

QuantDesk is in closed beta. Drop your email and we’ll email you when a spot opens.

Questions worth answering

Do you hold my funds?

No. Never. You authorise a trading agent (a sub-key) that QuantDesk can trade with — but cannot withdraw. You can revoke it at any time.

Can the system lose all my money?

Yes. Trading is inherently risky. QuantDesk copies certified lead wallets with your capital, and a certified lead can still lose. The system has hard kill switches (per-trade risk, daily loss, max drawdown) to prevent runaway bugs, but it cannot prevent adverse market moves or a lead going bad. You should only use capital you can afford to lose.

What about bugs?

There will be bugs. This is beta software. You accept that. Every AI decision runs shadow-first and is graded before it is trusted, and kill switches and hard caps exist to contain blast radius — but the risk is yours.

Is there a licence / regulation angle?

QuantDesk is software-as-a-service for self-directed traders. You connect your own wallet, you choose your strategy, you execute on venues you already use. We're working toward formal compliance as we grow; during beta it's disclosure-gated.

What does 'learn together' mean for my data?

Your positions, P&L, wallet, and chat are private and never shared with other users. What IS aggregated — anonymously, with tenant identity stripped — are the outcomes of AI judgments. If a pattern worked across 50 users, everyone benefits; if it failed, everyone avoids it next time.