Rusk: a small crew for a closer look

A market can move faster than the story around it. A rising chart attracts attention, a familiar ticker creates confidence, and a busy feed can look like proof. Rusk starts with a quieter question: what can the available data actually tell us? It brings three robot operators into a shared market research interface, with readable comparisons and source links close to every case.

The crew has three distinct roles. Scout reads changes in recorded trading activity. Anchor examines pool liquidity alongside market conditions. Tempo checks whether price movements agree across different time windows. Their readings sit beside each other so that differences remain visible. The characters give those roles a face, but the calculations underneath use explicit, deterministic rules. They are not connected language models, autonomous traders or a trained system for predicting winners.

The market lab begins with 18 curated Pump.fun-origin reference cases identified by their complete mint addresses. This is a reference sample, not a census of new launches or a ranking of everything on Pump.fun. The selection includes repeated tickers, making an important distinction visible: a symbol is a label, not a unique token identity. Each case links to its selected pool and external sources so that the reader can check what is being compared.

Quotes come from the documented DEX Screener API. Rusk selects the deepest valid pool among the indexed candidates returned for each exact base token on Solana; that coverage is not exhaustive. The lab shows market cap, liquidity, volume and additional descriptive measures, including short-window volume pace and liquidity relative to market cap. Missing market cap is not silently replaced with fully diluted valuation. Missing fields stay unavailable rather than becoming invented estimates.

Context matters more than a single impressive number. Market capitalization is not cash available to exit. Counted buys are not unique buyers. A faster volume window does not establish that demand is organic. Agreement between price windows describes the current snapshot rather than proving what happens next. Rusk exposes these distinctions through formulas, metric definitions, source details, filters and a CSV snapshot export. The page requests fresh data about every 60 seconds while it is visible, subject to provider availability.

The paper record turns a reading into something that can later be inspected. A directional observation records its token, source pool, initial quote, rule version and retrieval time. It can be checked against later quotes at one, six and 24 hours. A checkpoint settles only when a new quote from the original pool is retrieved at or after the scheduled time. If that happens late, the result is labelled accordingly; Rusk does not invent a price for an unobserved historical moment.

Those records are stored in the browser on the current device. They are not a shared, immutable public league, and they do not continue collecting quotes after the page is closed. A new browser starts without settled history. Matches, misses and pending checks remain part of the record. A direction match is a price-direction comparison, not a trading return: fees, fills and slippage are not simulated, and Rusk does not execute transactions, connect a wallet or hold funds.

The research notebook explains why these questions were selected. Graduation to another trading venue, liquidity that persists and a realized return are different outcomes. Studies of successful survivors cannot automatically reveal the chance that a fresh launch will succeed. Rusk uses published research to improve the questions and make limitations explicit. It does not present a calibrated success probability, a safety score or a validated strategy.

Off-white surfaces, black type and a clear blue accent keep the interface focused on the evidence. The robot crew makes the workspace recognizable without replacing the method with a mascot story. Rusk's proposition is straightforward: inspect the identity, compare the readings, keep the source and leave room for the result to disagree with the expectation.

640 words. Project description based on the current implementation. Source documentation: https://docs.dexscreener.com/api/reference

