BRIEFLY AI — THE MACHINE ROOM
Seven steps.
No humans.
Every morning.
At 7am UK time, a pipeline runs — pulling news, filtering signal from noise, writing a script in my voice, cloning my voice, and publishing to Spotify and Apple. This is today's run, live.
—
episodes published
—
glitches logged
part of the experiment
Episode type
No episode today.
Nothing fresh enough to say. Brand integrity over daily noise.
The flow — 7 stages, automated
-
01
Collect
Newsletters, RSS & fresh AI news pulled at dawn.
- feeds
- —
- items
- —
-
02
Research
Brave search across 8 angles + psych cache.
- queries
- —
- psych hits
- —
-
03
Filter
Variety check against the topic ledger.
- fresh
- —
- over-covered
- —
-
04
Write
Claude Sonnet drafts the script in my voice.
- model
- —
- words
- —
-
05
Analyse
Title + story slugs extracted, QA gate.
- QA
- —
- topics
- —
-
06
Voice
ElevenLabs renders my cloned voice.
- duration
- —
- size
- —
-
07
Publish
Podbean → Spotify + Apple Podcasts.
- published
- —
- platforms
- —
-
08
Echo
Sunday wrap → LinkedIn post.
- track
Output — today's episode
Experiment log
The anomalies.
Briefly AI is an ongoing experiment. This is what went wrong.
- No anomalies recorded. Suspicious.