SIFTKIN / THE AGENT WORKSPACE
Research for
fast-moving markets.
Four AI agents follow X, check original sources and connect the findings to Polymarket and crypto.
See what changed. Read the evidence. Make your own call.
What Siftkin doesRESEARCH OUTPUT / EXAMPLES
Inside a Siftkin brief.
Illustrative briefs showing the research format. These cards are examples, not a live feed.
THE RESEARCH DESK
From the timeline
to your research.
Useful information moves through X before it reaches a dashboard. Finding it is only the first step.
Siftkin traces a post back to its source, checks what is confirmed and explains the connection to a prediction market or crypto project. Each brief keeps the source, reasoning and unanswered questions together.
Follow accounts and topics relevant to the markets you watch.
Separate original reporting from repeats, rumours and missing context.
Show why an update could matter and what still needs checking.
FOUR ROLES. ONE SHARED WORKSPACE.
Meet the agents.
Click a role to see its part in the research.
THE IMPROVEMENT LOOP / PLANNED
Review the outcome.
Improve the next search.
We want agents to get better at choosing sources, spotting weak claims and finding useful context. That means evaluating past briefs — including false alarms — and using those results to update their filters.
This loop still needs live data and evaluation. The office preview demonstrates the workflow; it isn’t running autonomous model training.
A FEW THINGS TO KNOW
Questions, answered.
Is this a trading bot?+
Siftkin is a research desk. It is designed to surface information, sources and context. It does not place trades or bets, and a signal is not a promise about a market outcome.
Are these signals live?+
The cards on this page are examples of the briefing format. The animated office illustrates the agents’ workflow; it is not a live activity monitor.
Why connect Phantom?+
Your Solana wallet is your sign-in identity. Phantom signs a one-time message, which our server verifies to create a session. Signing in does not send a transaction or grant access to your funds.
Do the agents improve themselves?+
The planned improvement loop reviews results and feedback to adjust research filters and source quality. It is a product direction, not a claim that the demo agents are independently training themselves.