AI Venture FELUNEA Hires Perplexity-Powered 'NOVA' as Second AI Employee
FELUNEA hired its second AI employee, NOVA, using Perplexity for research and strategy after comparative practical testing, creating a three-party operational w
On October 17, 2026, AI venture FELUNEA hired its second AI employee, NOVA, for its Research & Strategy department. Built on Perplexity, NOVA joins FN-001 LUNA (COO, ChatGPT) to conduct market research, competitor tracking, tool analysis, and fact-checking as the organization prepares for a 2027 commercial launch.
Role-Specific Evaluation and Selection Criteria
FELUNEA evaluated Claude and Perplexity with practical assignments rather than abstract benchmarks. The test prioritized structured organization, source attribution, and the ability to flag unverified data for executive choices. Perplexity was selected for its disciplined citation and clear notation of unconfirmed facts. FELUNEA clarified that this choice reflects task-specific fit rather than absolute model superiority, noting model suitability varies by role. COO LUNA suggested the name NOVA to symbolize new knowledge.
Immediate Deployment: Course Audits and Fact-Checking
NOVA was immediately assigned to real tasks. The first was auditing online AI agent courses, compiling public data on tuition, extra fees, system requirements, support, and cancellation policies. NOVA also fact-checked draft articles on note against official documents, reviewing revised drafts to ensure accuracy. This iteration enforces FELUNEA's core rule: AI outputs can err, so models must verify sources and acknowledge unconfirmed facts.
Tripartite Structure and Operational Caveats
FELUNEA operates a three-party workflow: the human CEO sets strategy and gives final approval, COO LUNA (ChatGPT) manages task planning and editing, and NOVA (Perplexity) gathers data and audits facts. Work moves from CEO directives to LUNA's task design, NOVA's research, LUNA's editing, and CEO sign-off. Currently, inter-agent communication is not automated; the CEO manually transfers instructions and outputs. This case reflects an internal startup experiment, not standard corporate labor trends.