Why HypeTorch exists
I've spent my career in product management and data analytics, watching advanced metrics change how teams evaluate players. WAR, PER, expected goals — context-aware numbers that replaced counting stats.
Sports media never got that upgrade. Decisions about which athletes to sign, feature, or pay still run on follower counts and impressions. Those measure audience size. They don't measure influence.
HypeTorch measures what people actually say.
What it does
Every week, HypeTorch ingests podcast episodes, transcribes them, and identifies which entities were discussed, for how long, and in what context. Those signals become three scores:
Attention — who is being talked about right now
Attention efficiency — over- or under-valued relative to social reach
Divisiveness of coverage
Podcasts are the most honest long-form discussion available. Nobody optimises a two-hour conversation for engagement. What gets talked about is what people genuinely care about — and it shows up there before it shows up anywhere else.
Where the data comes from
If you're licensing this, you should know exactly what's underneath it.
Public RSS feeds only. No scraping behind logins, no terms-of-service violations.
Audio is transcribed and discarded. Source recordings are never retained or redistributed.
Transcripts stay private. They are retained in an internal store used for deduplication and re-scoring, and are never published or redistributed. What powers the scores is derived data — entity, timestamp, talk time, context.
Output is factual. Counts, durations, and scores. No source content is republished.
What it doesn't do
- —Podcasts and public web signals only. Not TikTok, Instagram, X, or Twitch.
- —English-language sources.
- —Measures attention, not brand sentiment.
- —Descriptive, not predictive. It tells you what's being said now, not what will happen next.
- —Coverage is bounded by the source panel. An entity discussed only on shows we don't index won't appear.
Coverage
Currently running weekly across Unrivaled and NBA.
New populations and new source panels are configuration, not development.
Who's behind it
HypeTorch is built and run by Zach Burma.
The pipeline has run unattended every week since January 2026.