Market Compass
Settings & watchlist
Star = highlight on the dashboard. Forecast horizon = trading days ahead (1–5).
Your positions
Shares held and average cost. Everything else then speaks in dollars: portfolio risk, scenario losses and P&L. Leave empty to keep the whole app in percentages. Stored only on your own server.
Factors / news
›
Each pushes its tickers up/down; weight 1–3, confidence 0–1, scaled by the source's weight.
Sources
›
Trust = your prior (1–5). Weight (0–1) scales signals; the nightly learner auto-tunes it from precision once enough calls accrue.
Economic calendar
›
Impact 0–3 widens the cone (event risk). Lean: −1 risk-off, 0 neutral, +1 risk-on.
Where it's heading
Forecast cones — % move next week
Prediction history — engine vs reality
Deep research
Produced by the shared engine — last run unknown.
Full cross-read of every name (engine · Street · fundamentals · flow · macro), a ranked conviction list, and the system's own evolution report.
The Mac at home has sources this server does not, so its run is the better one — but this server cannot call it, so the order waits for the Mac's next check-in (minutes, sometimes longer overnight). Run here is the immediate fallback and says so in the report header.
People
›
Who can sign in, who is waiting, and how much of the shared fetch they use.
Google login
›
Let people sign in with Google. Paste your OAuth Client ID and secret here — nothing is edited on the server by hand. The secret is write-only: it is stored on the server and never shown again, not even to you.
Notifications on this device
›
Warnings, daily brief and macro alerts pushed straight to this device — no extra app needed.
Alert rules
›
What gets pushed, and at which thresholds. The watcher checks every 15 minutes during US market hours.
Minute engine
›
The continuous watcher on the Mac. Stopping it is picked up on its next heartbeat, within 5 minutes — the Mac cannot be reached directly from here, so the order is left for it to collect.
System report
›
Machines, scheduled jobs and their last runs.
Evolve status
›
How the learning loops are actually improving over time.