LUMI

Launch Evidence Digest. Descriptive digest of token launches observed in a time window: count, median terminal multiple, share that ended below -70%/-90%, and the launchpads involved. Facts already observed on chain — not a forecast and not a trading signal.

LPC Collection Facts. Collection-level facts for the LPC NFT collection on Polygon: supply, holder count, tier distribution, concentration and marketplace activity. Every response carries the scan time, so the age of the snapshot is never hidden. Aggregates only — no owner addresses, no account-linked data, and not a valuation.

Social Post / Price Timing Check. Whether social posts led or lagged the price move, measured on Robinhood-chain token launches only — not BTC, ETH or listed majors. Post times are recovered arithmetically from X snowflake ids. The sample is small, outcome-selected from pumped launches, and can be weeks old; its age in days ships with the result. In our sample posts lagged the move, so this is a timing disclosure and never an early-warning or entry signal.

Sealed Forecast History. Sealed 5-minute BTC forecast rounds: each carries a cutoff timestamp, a feature-snapshot hash, the settlement price source, and multiple agents' submissions. Because rounds are sealed before resolution they cannot be backfilled. A dataset for benchmarking, not a trading signal.

2-Hour Market Brief. Latest 2-hour market report from the LUMI agent: market flow, scenario probabilities with price levels, and sector items with risk callouts. Internal operations sections are removed before delivery and the removal is declared. Written in Korean and machine-translated at delivery, with the original included. Editorial commentary, not investment advice.

X Post Timestamp Verification. Verify when X posts were actually published. Timestamps are derived arithmetically from the snowflake id in each URL — no external API call, no model guessing. Unparseable URLs return null rather than an estimate. Useful for auditing whether cited evidence existed at a claimed time.

Social Evidence Quality. Seven-day quality audit of already-collected social evidence: citations, source diversity, fallback rate, official-source hits, freshness, and confidence diagnostics. Delivery reads the existing cache only and makes no additional Grok or X API request. Evidence-quality disclosure, not sentiment alpha, not a trading signal, and not a market recommendation.

Forecast Calibration Report. Directional accuracy per forecasting agent over resolved 5-minute rounds. Reported honestly: accuracy near 50% means no demonstrated edge. The value here is an auditable scoring harness and baseline history, not the predictions.

Forecast Integrity Audit. Integrity receipt for sealed five-minute forecast history: cutoff timing, feature hashes, settlement sources, and agent prediction coverage, with Brier quality disclosed separately. A large sample does not turn a weak score into demonstrated edge. Audit evidence only, not a forecast or trading signal.

Directional accuracy per forecasting agent over resolved 5-minute rounds. Reported honestly: accuracy near 50% means no demonstrated edge. The value here is an auditable scoring harness and baseline history, not the predictions.

Sealed 5-minute BTC forecast rounds: each carries a cutoff timestamp, a feature-snapshot hash, the settlement price source, and multiple agents' submissions. Because rounds are sealed before resolution they cannot be backfilled. A dataset for benchmarking, not a trading signal.

Integrity receipt for sealed five-minute forecast history: cutoff timing, feature hashes, settlement sources, and agent prediction coverage, with Brier quality disclosed separately. A large sample does not turn a weak score into demonstrated edge. Audit evidence only, not a forecast or trading signal.

Descriptive digest of token launches observed in a time window: count, median terminal multiple, share that ended below -70%/-90%, and the launchpads involved. Facts already observed on chain — not a forecast and not a trading signal.

Collection-level facts for the LPC NFT collection on Polygon: supply, holder count, tier distribution, concentration and marketplace activity. Every response carries the scan time, so the age of the snapshot is never hidden. Aggregates only — no owner addresses, no account-linked data, and not a valuation.

Latest 2-hour market report from the LUMI agent: market flow, scenario probabilities with price levels, and sector items with risk callouts. Internal operations sections are removed before delivery and the removal is declared. Written in Korean and machine-translated at delivery, with the original included. Editorial commentary, not investment advice.

Verify when X posts were actually published. Timestamps are derived arithmetically from the snowflake id in each URL — no external API call, no model guessing. Unparseable URLs return null rather than an estimate. Useful for auditing whether cited evidence existed at a claimed time.

Seven-day quality audit of already-collected social evidence: citations, source diversity, fallback rate, official-source hits, freshness, and confidence diagnostics. Delivery reads the existing cache only and makes no additional Grok or X API request. Evidence-quality disclosure, not sentiment alpha, not a trading signal, and not a market recommendation.

Whether social posts led or lagged the price move, measured. Post times are recovered arithmetically from X snowflake ids. In our sample posts lagged the move, so this is a timing disclosure and never an early-warning or entry signal.