Sign in to control your local listening desk, review the transcript, and follow markets as the speaker changes topics.
This installation controls one team laptop and its connected trading account. Approved operators should keep that laptop worker running while listening or replaying a speech.
Get the MarketPulse project on the team laptop, then run pip install -r requirements.txt in its folder.
Put each API key in its matching local file in that folder: gemapi.txt (Gemini) and elevenapi.txt (ElevenLabs) are needed for live listening and scoring.
Add hostinger_url.txt with this site's URL and hostinger_token.txt with the site's private upload token. The local worker uses them to publish updates.
For optional features, add fredapi.txt, socialcrawlapi.txt, backboardapi.txt, and tigerdb.txt. Venue credentials belong in kalshikey.txt with privkey.txt, or polymarketkey.txt with polymarketsecret.txt.
Run python control_worker.py and return here. Start in Dry run mode.
Keep keys on the team laptop. Do not paste them into this page, upload them to the website, or commit them to Git. Auth0 sign-in does not supply local API keys. Another team needs its own isolated deployment and worker to use its own account.
MarketPulse
Live Fed-speech surprise detection → prediction-market orders
—connecting…
Control room
Checking operator controls…
Press conferences and speeches by Warsh and Powell, plus presidential remarks on the economy. The speaker baseline switches automatically.
50 MB max. Pick a playback speed below.
YouTube (live or recorded) and federalreserve.gov.
Try: “Inflation is still far too high. We are prepared to raise rates again in October.”
Replays default to 10×, recordings to 1×.
The Kalshi + Polymarket market recommenders start automatically with every session.
Limits: $10/order, $50/day, kill switch on the laptop. LIVE requires approved Auth0 sign-in, typing LIVE, and local worker opt-in. Analyze always uses dry run.
Status
—
Latest stance
—
Surprises
0
Signal → order
—
words heard to orders sent
Surprise vs. the speaker's usual stance (z-score)
Above +2 = hawkish surprise, below −2 = dovish surprise. Each dot is a passage of speech.
Transcript
Waiting for speech…
Recent order attempts
No order attempts yet.
Ready to trade
Waiting for the speaker…
—
Live market recommenders
Not running.
Markets in view
Markets follow the latest speaker topic. Every real order requires a fresh exchange quote, a short-lived preview, and a separate confirmation. This desk uses one team laptop and trading account.
Waiting for markets…
Active orders
Sign in with an approved Auth0 account to view private exchange data.
No exchange snapshot yet.
Holdings
No exchange snapshot yet.
Recent order attempts
No order attempts yet.
Ask the desk
Answers come from Backboard memory: every surprise signal and the orders it triggered.
Sign in above to ask questions.
What the desk remembers
Sign in to browse memories.
How surprising is this, for this speaker?
Uses the speaker, trading mode, contracts and venues chosen in the control room.
Sign in above to run an analysis.
What the crowd expects (Bluesky + Reddit)
Recent posts about the Fed, scored on the same −1 dovish / +1 hawkish scale as the speakers. A speaker surprise that runs against the crowd is a bigger repricing.
Loading…
Macro backdrop (FRED)
Latest data from the St. Louis Fed. Gemini sees these numbers when it scores each passage, and each speaker's usual stance is shifted by how much core inflation and unemployment have moved since their baseline was recorded.
Loading…
Replay
Pick a speech: browse its pre-scored answers here, or press Play through the live desk to stream it through the real pipeline (transcript, surprises, markets and orders in the chosen mode) on the Live tab.
Replay tab: precomputed stance-only results, scored answer by answer. Control room replay: feeds a saved transcript through the timed live desk, updating the Live tab and showing order attempts in the selected trading mode. Start with Dry run.
Scored answer by answer. |z| ≥ 2 is a surprise.
Loading…
Signals (Tiger Data)
Loading…
Orders
Loading…
How it works
0 · ContextFRED macro data and Bluesky/Reddit crowd sentiment frame each surprise.
1 · ListenElevenLabs Scribe v2 Realtime turns live audio into text, sentence by sentence.
2 · ScoreGemini rates each passage −1 (dovish) to +1 (hawkish), compared with this speaker's own past answers.
3 · DecideA z-score of ±2 or more is a surprise. Markets were picked before the speech, so there's no search delay.
4 · TradeKalshi + Polymarket orders within $10/order, $50/day limits and a kill switch.
5 · ProveEach call is hashed onto Solana, stored in Tiger Data, and remembered by Backboard.
