| name | news-heat-check-ops |
| description | Measure the current virality, velocity, and editorial usefulness of a news story, product launch, model release, funding event, or trend. Use when the user asks for a heat score, asks how hot/viral/trending something is, wants to compare competing news waves, decide whether a topic is worth covering now, distinguish mass attention from niche audience fit, or audit why a radar missed a high-signal canonical announcement. |
| memory_tags | ["domain:research","workflow:news-heat-check","skill_role:evaluator","repo_boundary:tools","inputs:current-sources","outputs:heat-score","risk:medium"] |
News Heat Check Ops
Produce a source-backed, point-in-time heat comparison without confusing a
large account's owned distribution with genuine breakout.
Inputs
- One or more topics, announcements, URLs, or candidate stories.
- Optional: geography, audience, editorial lane, comparison topic, and time
window. Default to the last 24 hours and the user's apparent audience.
- Optional: a historical breakout to use as a calibration reference.
Workflow
- Define the snapshot.
- Record the timezone, snapshot time, and comparison window.
- Treat counters as point-in-time observations that will continue moving.
- Build the candidate universe.
- For a multi-story radar or open-ended sweep, read
candidate-discovery.md and complete
its coverage gate before ranking.
- Search primary announcement surfaces and independent discovery surfaces;
do not begin from the personalized feed alone.
- For a single user-supplied topic, skip the broad universe sweep but still
resolve the primary announcement.
- Resolve story clusters.
- Group aliases, product names, company names, and duplicate submissions
into one story cluster.
- Keep similarly named but distinct stories separate.
- Treat a secondary post that references a launch as a candidate-discovery
event. Resolve it back to the canonical announcement even when the
official account or post did not appear in the user's feed.
- Treat personalized feeds as one discovery surface, not the candidate
universe.
- Find the primary evidence.
- Start with official announcement posts/pages and direct discussion links.
- Before ranking, audit every launch cluster for the company, founder, or
official product announcement that best represents the primary event.
- Browse current sources; do not answer a live heat question from memory.
- Measure the available surfaces.
- Social: views, engagement, publication time, author followers, secondary
posts, and post-count velocity.
- Discussion: Hacker News points/comments, Reddit conversations, replies,
quotes, and substantive follow-on analysis.
- Video/search: YouTube views per hour, short-form crossover, Google search
lift, and news coverage when available.
- Developer signals: GitHub star/event velocity only when the story has a
relevant repository.
- Calculate comparable signals.
- Views/hour = observed views / hours since publication.
- Views/follower = observed views / current author followers. Label current
followers as a conservative proxy when historical followers are missing.
- Separate one large owned post from distributed cross-account saturation.
- Score two different questions.
attention_heat: velocity, follower-adjusted breakout, cross-platform
spread, discussion depth, and acceleration.audience_fit: usefulness, demoability, novelty, relevance, and the
availability of a distinctive angle for the stated audience.- Read scoring.md for the rubric and evidence rules.
- Optionally run
scripts/score_heat.py after assigning component scores.
- Make the editorial call.
- Label each cluster
watch, rising, hot, peaking, or cooling.
- Recommend
ignore, mention, single segment, or drop-everything.
- Explain the strongest evidence, the most important caveat, and what would
change the decision.
Priority Order
When time is constrained:
- Candidate coverage gate and canonical primary resolution.
- Primary announcement metrics and timestamp.
- Views/hour and follower-adjusted breakout.
- Hacker News/Reddit discussion depth.
- Cross-account and cross-platform spread.
- Search, video, GitHub, and paid-tool confirmation.
Do not spend the entire budget filling every source. Missing data is a labeled
gap, not a zero.
Output Contract
Return:
- Snapshot time and scope.
- A side-by-side table with raw metrics,
attention_heat, audience_fit,
stage, action, and confidence.
- A short explanation of owned reach versus breakout reach.
- A verdict for each story and the best timely angle, if one exists.
- Direct links beside the claims they support.
- Source gaps and historical-counter caveats.
- For multi-story sweeps, a compact coverage audit listing high-signal
candidates considered but omitted and the reason for each omission.
When writing to a radar or research artifact, put this comparison before long
source inventories.
Guardrails
- For routine or automated X reads, use the Bird-style browser-session path
(
AUTH_TOKEN + CT0) through XBot or IgnitionBot. Do not use an X developer
bearer token or the official recent-search endpoint, and never fall back to
a billable official read when session search fails. Record stale session or
GraphQL query IDs as a source gap.
- Do not call a story viral from raw views alone.
- Do not treat a personalized feed or saved seed list as the market universe.
- Do not substitute a low-follower/high-view outlier scan for event-level news
detection. Account-relative breakout is one scoring component, not the
candidate generator or the complete heat score.
- Do not leave a secondary launch reaction unclustered when the canonical
announcement can be resolved.
- Do not compare lifetime counters as if they were equal-age observations.
- Do not infer historical velocity from a current cumulative counter.
- Do not treat press volume, followers, downloads, or GitHub stars as proof of
adoption, revenue, or retention.
- Keep evidence and interpretation visibly separate.
- Treat scores as editorial heuristics until calibrated against actual content
performance.
