Methodology

What we monitor

Mentioned currently monitors eight public investing and crypto forum communities, rotating four two-community groups on scheduled runs. One group is normally revisited every two hours. Job delays, source availability, and rate limits can stretch that interval, so the displayed snapshot time—not the schedule—is the freshness authority. The monitored set and access paths can change as sources evolve. The product is the aggregate rollup, not a map to individual posts. New posts are pulled through permitted public or platform-provided access paths.

How posts are analyzed

Each post runs through a structured pipeline. Tickers are extracted and normalized, then screened against a broad false-positive list so ordinary words, indices, currencies, and forum slang do not dominate the feed. We surface recurring topic keywords so briefs stay anchored to the aggregate discussion pattern, not a one-word market call.

Premium chatter briefs are available for qualifying leading tickers. They use only the aggregate mention count and recurring themes shown by this pipeline—never raw post text. The server may use a constrained language model or a deterministic aggregate fallback; either way, the output stays source-agnostic and explicitly not financial advice.

Deduplication and noise controls

Every source post is stored under its source ID, so ingesting the same item again updates the existing row instead of adding another count. Within a post, a ticker counts at most once even if its symbol appears repeatedly. Before ranking, uppercase tokens also pass a conservative false-positive list that removes common words, forum shorthand, currencies, and other frequent non-ticker collisions.

Mentioned does not currently claim to identify accounts, botnets, or coordinated campaigns. Promotional or organized posts can still be present in the aggregate. That limitation is shown plainly rather than assigning an unverified “clean” label to the count.

How trending is calculated

The current public feed ranks tickers by aggregate mention count inside the latest stored window and attaches repeated keyword clusters. Snapshot freshness is shown in the app so users can distinguish current, stored, and unavailable states. More advanced baseline math can be added later, but the public page does not imply a formula that is not active in production.

What we report

Briefs summarize the most-discussed names and recurring topic keywords when the product has enough measured data to support them. Ranked updates list aggregate ticker counts, repeated keywords, and the feed window. The system never adds buy or sell language. It does not attach price targets or personal guidance.

What we do not do

• We do not issue buy or sell recommendations
• We do not generate price targets
• We do not give personal advice
• We do not trade ahead of subscribers
• We do not sell user data
• We do not track which tickers individual subscribers privately watch in order to build positions

Limitations

The agent is useful for observing mention trends. It does not predict which trends become price moves. We do not claim it does. Public chatter can be promotional, coordinated, sarcastic, mistaken, or manipulated, and the current rollup does not reliably classify those motives. Output quality depends on what is publicly readable. Private chats, paywalled rooms, and offline flow are invisible.

How this page changes

When methodology changes, we update this page and keep material implementation changes versioned in the repo that powers the agent so there is a paper trail.