Hot Trend Lab Publishing Protocol

A column works when it can be refreshed without reinventing the analysis each week. This protocol turns the notebooks into a weekly editorial loop.

Weekly loop

Day Action Output
Monday scan refresh source cards and run the low-cost notebooks candidate table with coverage and residual ranks
Midweek verify inspect query wording, API limits, cache dates, and suspicious spikes short list of publishable charts
Friday publish write the article around one source card and one clear judgment source table, figure, boundary paragraph, notebook link

Residual trigger rules

Use residuals as prompts for closer reading, not as final conclusions.

Trigger What to check before writing Likely article form
Category residual spike conference deadline, query drift, duplicate cross-listing arXiv research pulse
Hugging Face snapshot jump launch date, model family, benchmark context, repeated snapshots open-model pulse
GitHub star velocity burst pagination coverage, token limit, media or demo event developer attention pulse
Wikimedia attention spike article title, news event, source project, decay speed public-attention note
Crypto residual event BTC/ETH move, stablecoin context, venue or macro event market-structure note
AI infrastructure basket move basket definition, sector or market benchmark, earnings calendar market-proxy note

Publish/no-publish thresholds

Gate Publish when Do not publish when
Minimum observations the chart has enough points for the selected method and period warm-up or cool-down dominates the event table
Freshness access date matches the article date or stated cache date cache date is unclear
Residual threshold residual event is large enough to survive chart inspection and source context the event is a first/last-window artifact
Source failure failure is visible and the article can be framed as a source limitation missing source is silently replaced
Artifact checklist source card, figure, summary table, boundary paragraph, and notebook link are present any artifact is missing

Article contract

Every article includes:

  1. a source card with query, access date, coverage, and limits;
  2. one decomposition summary table or figure;
  3. a residual event table when the story is event-driven;
  4. two or three sentences explaining the chart axes and signal;
  5. a boundary sentence naming what the chart does not measure;
  6. a rendered notebook transcript link.

Cadence map

Cadence Article Notebook Publish when
Weekly arXiv AI category pulse 01 arXiv category pulse source coverage is complete and deadline context is checked
Weekly AI agent / coding-agent research pulse 02 arXiv agent research pulse query terms are stable and a residual event has a clear topic
Weekly Hugging Face open-model pulse 03 Hugging Face open-model pulse snapshot date and ranking scope are explicit
Weekly GitHub AI-agent star velocity 04 GitHub AI-agent star velocity coverage window and pagination limits are shown
Monthly Wikimedia attention and hype decay 05 Wikipedia attention hype decay article set and date window are stable
Event-driven crypto and stablecoin liquidity pulse 06 crypto stablecoin liquidity pulse a residual event has market-structure context
Earnings / macro weeks AI infrastructure market pulse 07 AI infrastructure market pulse basket definition and benchmark comparison are included