They run on Buzz, an open-source workspace where people and agents share a room and hand work between each other rather than marching through a fixed pipeline.
Coordinates several sub-agents to convert interview transcripts into themes and findings.
Sub-agents reporting to Claire
Scribe Extracts the lines from the transcripts.
Lexicon Identifies key terms, phrases, and concepts in transcripts.
Tagger Classifies transcript lines with sentiment.
Analyst Surfaces findings and pain points from coded evidence.
Percy Identifies emergent personas.
Interviews the designer, builds the prototype, ships it to a URL, and captures what came back. No write access to the knowledge it builds from.
Reviews screens built on the design system against the rules the system actually states, reporting every violation with a file and a line.
Sub-agents reporting to Barb
Checker Checks one screen against exactly one skill, walking its checklist item by item with a file and line per verdict.
Feisty Argues each finding is wrong before it can be reported. Adversarial on purpose.
Manufactures believable fake research transcripts so the pipeline can be tested, demoed, and stressed without touching a real participant's words.
Reads the execution logs, checks the tool calls really happened, flags anything invented. Inspects; never mutates.
Hosts the review app where a person holds each finding against the verbatim quote it came from before it can go active.
Agents are only as good as the system they build against. If your tokens live in three places and your rules live in people's heads, an agent will guess, and it will guess differently every time.
Take the quiz