Collaborate on the research
Market Synthesis is an open PhD research project. It improves fastest when people who understand markets tell it where it went wrong. There is no product to buy and nothing to sell — just a research system whose predictions are published openly so they can be examined and improved.
Research output — not investment advice. Nothing here is financial advice, a recommendation, or a solicitation. The author is not a financial adviser. Predictions are experimental and frequently wrong.
The most useful feedback
When the system makes a call and the market does something else, the reason is gold. The kinds of feedback that directly shape the next run:
- Missed events. “This prediction broke because event X happened that the debate never saw.” → tells us where to add news enrichment.
- Missed adversarial angles. “The Defender should have argued Y.” → improves the category-specific debate prompts.
- Wrong instrument / direction mapping. “The thesis was right but the tradeable expression was wrong.” → improves execution routing.
- Data-quality gaps. “Your price/news source is stale or wrong for this asset.” → improves the data layer.
How to contribute
- Discuss a specific gap: open any entry in the detected-gaps log and use the discussion thread at the bottom of the page.
- General discussion / suggestions: the community discussions.
Discussion runs on GitHub Discussions (free). Reading requires nothing; posting a comment requires a free GitHub account — a deliberately low bar that keeps contributors identifiable and the signal-to-noise high.
What I’m researching
This is a PhD project at RMIT University: an adversarial multi-agent LLM framework for macro-financial regime detection. The open questions I’m actively working on:
- Where does the framework’s edge actually live? (Current evidence: slow structural narrative-vs-reality gaps — not fast reaction to public news or reflexive markets like crypto.)
- How much does richer news enrichment improve detection?
- Can community feedback measurably reduce the system’s error rate over time?
If you’re an academic or practitioner who wants to collaborate on the research itself (not a trading relationship), the discussions above are the place to start the conversation.
An open PhD research project — Ben Lvovsky, RMIT University. Not investment advice.