KandaQuantum Releases Fuga 5000, a Swarm Orchestrator for 5,000 Agents a Day
Fuga 5000 sells the orchestration layer rather than the models, piggybacking on subscriptions developers already pay for; the company's own measurement puts a workload at about 47 dollars a day against roughly 475 dollars at API metered rates.
Reporting from 1 source: ASCII.jp.
KandaQuantum has released Fuga 5000, an orchestrator that coordinates swarms of AI agents for developers. The company says it manages up to 5,000 agents a day from one Mac, connecting to existing Claude Code, Codex, Gemini, Grok, and Kimi subscriptions. In a test on October 16, 2026, on a single MacBook Pro, it distributed 5,441 agents across 349 tasks. Fuga 5000 is macOS-only, and existing users already have access.
Fuga 5000 splits the work across five layers, from a request-handling front end down to sub-agents beneath each main agent. Goal-level PM agents sit in the second layer and do not touch task detail; approval steps return to the user through the first layer. Assignment is automatic: a ledger pulls in current web benchmarks, and a pie chart sets how the vendors are split, after which tasks land on a Gantt chart that runs from planning through parallel implementation, integration, reporting, independent inspection, and merge into the base branch.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.