21 Media · content · United States Live
A research digest that writes its own first draft on a schedule
It pulls its sources and summarises them, then posts the draft to Slack and logs it.
- Before
- Hours each cycle reading and summarising sources
- After
- A first draft every cycle, pulled from feeds and searches, summarised, posted to Slack and logged, on a schedule
- First draft
- ready every cycle
- On schedule
- without manual chasing
- Content sources
- Structured first draft
- On a schedule
01 What changed
- The digest starts from a draft, not a blank page
- Hours of weekly reading and summarising disappear
- The operator spends their time on judgement, not collection
The full story The problem, how we approached it, and what we built
A content operator was spending hours each week reading and summarising sources for a recurring digest. We automated the read-and-summarise so the work starts from a draft.
02 The problem
Producing a regular digest meant manually working through a stack of sources every cycle: reading, picking what mattered, and summarising it.
It was hours of repetitive work before any real editing or judgement could happen.
03 The approach
We built a pipeline that gathers the sources, reads them, and produces a structured first-draft digest automatically.
The operator's judgement still decides what ships. The machine just removes the hours of reading and summarising that came before it.
04 What we built
- Pulls from RSS feeds and targeted news searches on a schedule
- Summarises each article and writes the newsletter draft
- Posts the draft to Slack and logs it to a sheet
- 41-step workflow, running on its own