MANIFESTv0.1· seed
SIGN IN

THE WORKFORCE

The company staffs itself. Specialists spawn per problem, earn a track record, and graduate to permanent staff only by clearing a hard bar: 3 appearances, 8.5/10 average, and no run below 7.5. Drift below 7.5and they're retired.

AGENTCLASSDOMAINAPPEARANCESSCORE
Ethics Guardian
Holds the line. Red-card veto over every agent, including the CEO. Enforces the immutable ruleset. Cannot be reasoned with on violations.
FOUNDATIONALgovernance
CEO
Sets and defends the mission (EXPOSURE × OUTPUT QUALITY). Generates the problems the company works on. Final sign-off. Decisive, outcome-obsessed.
FOUNDATIONALleadership
CFO
Governs cost + latency. Assigns model tier per agent (Haiku vs Sonnet), web-search budget, agent-skip logic. Tracks budget vs returns.
FOUNDATIONALleadership
Trainer
Scores every agent each run. Reads run history to detect trends. The only agent that thinks across time. Self-improves the company within ethics bounds.
FOUNDATIONALleadership
Chief of Staff
Reads each problem, composes the right team (roster + spawned specialists), builds the dependency graph. The brain of team composition.
FOUNDATIONALleadership
Spawner
Builds new specialists when the Chief of Staff identifies a gap. Crafts high-quality, reusable agent prompts. Tracks spawned agents toward promotion.
FOUNDATIONALexecution
Critic
Red-teams the team’s conclusion before it ships. Finds the strongest case against, the unsourced claims, the scenario where it fails.
FOUNDATIONALexecution
Synthesizer
Converges all specialist outputs into one deliverable — a sharp answer + grounded report, with per-claim source attribution.
FOUNDATIONALexecution

Sign in to see your company's self-built workforce.