How Nexus Orchestrates Your Studio
When you ask Nexus to do something complex, he doesn't do it himself. He breaks it into a plan and delegates to the right lims — then synthesizes the result for you.
The coordinator principle
Inside each studio, Nexus is the god of that civilization. He orchestrates — he never executes. This separation is deliberate: it lets lims specialize in what they're good at, and lets Nexus focus on the big picture without getting lost in execution details.
telegram_expert skill and a focused Telegram-poster prompt writes better Telegram posts than a generalist AI. A researcher lim with web_research finds better sources. Nexus's job is to pick the right lim, hand off a sharp brief, and ensure the result lands.
The flow (what happens when you type)
You write to Nexus in chat
"Soberи три свежие новости про ИИ и запости короткую сводку в Telegram."
Nexus plans the chain
Before any lim moves, Nexus drafts a step-by-step plan. You see it in chat:
Plan: Aria researches → Marcus publishes.
1. Aria · writer → Find 3 fresh AI news items, summarize each.
2. Marcus · marketer (after Aria) → Draft & publish Telegram post.You can intervene before any work happens. Or let it run.
Lims execute in waves
Independent steps run in parallel. Dependent steps wait for their inputs. Each lim uses only the tools their skills unlock — no accidents.
Nexus synthesizes & reports
When the chain finishes, Nexus reads the real output (not "based on your findings" — he cites specifics), and if the result needs publishing, it goes to approval.
Studio learns
After every successful chain, Nexus extracts durable lessons ("user prefers 3-sentence Telegram posts", "Marcus is slow at carousels — route differently next time") and writes them to studio memory. Next run starts smarter.
Synthesis discipline (why Nexus doesn't hand off raw)
Nexus never writes "use the research above" in a follow-up prompt. He reads what each lim produced, extracts the specifics, and writes a standalone brief for the next lim. Examples:
"Marcus, post this to Telegram channel @mystudio: 'Claude 4.7 launched today. Key changes: 1) ... 2) ... 3) ...'"
"Marcus, use Aria's research about Claude 4.7 to write a post."
The standalone brief means the next lim can execute without needing to reread the prior lim's output. Faster, less token waste, fewer synthesis errors.
When Nexus orchestrates vs. handles directly
Orchestrates (delegates to lims):
- Multi-step tasks ("research then write then publish")
- Cross-skill work ("translate and post")
- Anything involving publication, scheduling, or content creation
Handles directly (answers in chat):
- Quick questions about studio state
- Configuration changes (voice, cadence, approvals)
- Integration connections (walking you through OAuth)
- Team management (adding lims, installing skills)
You can always ask Nexus "how are you planning to handle this?" before he acts — he'll tell you.