Journal first
Write what you saw and heard. Attach audio, photos, people, place, and cultural significance without opening a forty-tab entry form.
Why Glossa
Built for the work before the dictionary.
Most language tools begin when analysis begins. Glossa begins earlier — when somebody notices a phrase, records a voice, photographs an object, or writes down why an event matters.
OCM in the paragraph
Tag a sentence with the Outline of Cultural Materials. Browse the tree later and see exactly which paragraphs document a topic.
People, not speaker codes
Participants and speakers are first-class. Mention a person, flag a voice, map kinship, and keep GEDCOM as interchange — not identity.
Transcribe with your hands on the keys
Waveform, rewind-on-resume, pitch-safe speed, speaker and timestamp chips, and a focused transcript editor built around the typing loop.
Analysis stays linked
Segment speech, then analyze one utterance at a time: word → morph → gloss → optional lexicon sense. No spreadsheet over a whole recording.
Follow, cache, fork
See teammates on the LAN and cache their journal feed read-only. Add their audio to your library when you want to analyze your own copy.
The principle
A feed, not a merged project.
Shared mutable databases turn fieldwork into conflict resolution. Glossa keeps authorship visible: follow a teammate, cache what they publish, read their annotations, and fork the recording when you want to do your own work. Their rows stay theirs.
0required cloud accounts
1portable project folder
796OCM headings included
sharing.glossa
peer appears on LAN
↓
follow → cache journal + media
↓
read their post and annotations
↓
fork audio → new local media id
↓
segment · transcribe · link senses
// no merged lexicon
// no edits to somebody else's rows
// no server required to analyze a language
The interface
One machine. Three bands.
A collapsible section drawer, one or two browse columns, and the editor where the work happens. The flavor changes. The window does not reinvent itself.
TODAY · KALO VILLAGE · AUTOSAVED
Funeral songs at the west house
Mara led the second song after sunset. @mara called it a road for the name to leave.
The women answered in two groups. 760 Death 536 Music
Open in Corpus → transcribe → mark utterances → analyze
Journal
TipTap blocks for prose, image, and audio. Hashtags become durable OCM chips; @ mentions become people.
Corpus
CodeMirror and wavesurfer for titled texts and recordings. Timestamps live in the transcript, not in a mystery sidecar.
Lexicon
A smaller FLEx-shaped model: entries, senses, examples, relations, and explicit LIFT interchange.
funeral-song.glossa
<!-- 00:42.7 --> <!-- @mara -->
Kála na náwi, kála na…
<!-- 00:48.2 -->
The west group answers.
F5 play / pause
F6 rewind
↵↵ drop timestamp
⌘S already saved
The desk pass
Chunk before you gloss.
Speech begins with time. Mark an utterance on the waveform, transcribe it without leaving the keyboard, then open analysis only inside that interval. Link a morph to a lexicon sense when you know it. Leave it unlinked when you do not.
F5typing loop
WAVmedia stays on disk
LIFTbridge, not identity
How it works
Field first. Desk when it helps.
Collecting and analyzing are different jobs. Glossa gives each the surface it deserves.
Capture the context
Write a short field post. Add a voice, photo, place, speaker, and OCM tags while the moment is still specific.
Find what the team knows
Browse the journal by timeline, person, or OCM tree. Coverage heat makes documented topics and cultural gaps visible.
Analyze what you own
Open media in Corpus, segment and transcribe, then connect words and morphs to senses in your lexicon.
In active development
The field notebook that grows a language.
The first desktop bake is underway. No cloud account, no Windows VM, no pretending a phone is a linguistics workstation.