A1 ยท Reading: community notes
Read each community on its own terms
Before it looks for patterns across communities, the aggregator reads each one and checks what the evidence it was given can support. On this site, a community is one subreddit read as a case; the run's files call it a room.
Where this stage starts
Community coding has produced first-order codes tied to participants' words, second-order themes that group those codes, and a written account of each community. Reply trees record who answered whom. The aggregator receives this coded collection as the community coders sealed it, so checking the collection's coverage is part of the analysis.
What the aggregator does
First, the aggregator checks the package: whether files and counts agree, whether dates make sense, and whether missing material or work on other platforms could change an interpretation. A community that tried and failed to organize stays in the analysis. Dropping it because it left no lasting artwork would hide part of the phenomenon.
The aggregator declares a reading order and its expectations, then reads each community as a whole. Participants' words come first, the community coder's account second and canvas information last. This order separates what the record shows from what another agent has already interpreted.
Each community gets a note on its aims, its sequence of events, its uncertainties and its last observed coordination. A later claim cites the source material, and the note only gives it context. Silence at the end of a thread leaves open whether organizing ended, above all when members moved to another platform.
What the next stage receives
B1 receives a reading of every community, a declared analysis set and the recorded uncertainties. From these it builds categories that work across cases.
What the output files contain
Markdown files hold prose; JSONL files hold one structured record per line.
Community reading notesrooms/<sub>.md, <sub>, order-A1.txt
rooms/<sub>.md records each community's story and
uncertainties. <sub> stands for its subreddit name.
order-A1.txt records the order of reading, so later readers
can see how the aggregator met the cases.
Words, patterns and possible connectionslexicon-A1.md, running-A1.md, leads-A1.jsonl
lexicon-A1.md keeps participants' vocabulary with source
references. running-A1.md collects observations across
readings. leads-A1.jsonl records possible connections
between communities that still need checking.
The planned comparison and exclusionscontrast.md, dropped.jsonl
contrast.md declares the early-versus-late comparison
and the expectations written before the reading.
dropped.jsonl names each exclusion and its reason; an empty
file means none were recorded.
How the end of organizing is observedcoordinating-rule-A1.md, organizing_end-A1.jsonl
coordinating-rule-A1.md sets out how coordination is
told apart from recounting or joking.
organizing_end-A1.jsonl records the last observed
coordination and what remained, including cases where the record is
silent.
Checks, departures and the handoffnotes-A1.md, defects.jsonl, departures.jsonl, run_manifest.json, handback-A1.md, _head.md
notes-A1.md explains decisions and limits.
defects.jsonl records problems with the data or tools;
departures.jsonl records departures from the protocol.
run_manifest.json records inputs and model settings.
handback-A1.md summarizes the handoff, and
_head.md records the current position for continuation.
Full stage description
Read the complete technical card for exact tasks, reading boundaries, checks and closing rules. The explanation above is a guide; the card holds the instructions.
Show how the stage is divided into its 3 legs
A1a: checks and screen. Repeats: no. Interview point: -.A1b: reading block. Repeats: yes. Interview point: -.A1c: lexicon and analysed set. Repeats: no. Interview point: end.
Source versions and hashes identify the exact technical documents reproduced here.