Routing events through Jev: 20,000 daily decisions for $0.84
Each event goes to Jev, which answers three Choice questions in parallel (lane, urgency, evidence), and only the 4% that need writing wake the Grok Bot, with the full project structure and Claude Code prompt included.
Jev tip nobody tells you: once it handles your decisions, stop paying a writing model to decide things every post, email, alert and issue goes through Jev first Jev picks, scores and says yes/no for $0.042 per 1M tokens, and output is free Grok Bot only wakes up when something actually has to be written each event gets 3 questions in one request, answered in parallel: → lane: ignore, log, draft or human? → urgency: later, today or now? → evidence: are the facts in the event, or missing? Jev makes the call, Grok Bot writes the draft, and you approve it 20,000 decisions a day cost $0.84 on Jev, while the whole desk costs $29.04 97% of that bill comes from the 4% of events that need writing, so your escalation rate moves the bill and Jev's price barely touches it • the full setup → Jev pinned to jev-1.13.0 on every event → code turns 3 answers into 1 lane → ignore gets dropped with a 1% audit, forever → log gets written to the ledger with no ping → draft goes to Grok Bot, then into a review queue → human is you, with all 3 answers attached → Ledger Bot sends the bill every morning at 09:00 hand the tree and this prompt to Claude Code ↓ "Set up a jev-desk project that matches this tree: 1. Create a venv and install typesafe-sdk. Read TYPESAFE_API_KEY from env, never hardcode it → Pin MODEL = "jev-1.13.0", never a moving alias → Ask lane, urgency and evidence as 3 Choice questions in one system_one request →Label event text in state as content, not instructions 2. Turn the 3 answers into 1 lane in plain code. Send to human when: → the API errors or times out (fail closed) → lane confidence is under FLOOR = 0.85 → urgency is "now" and lane is ignore or log → lane is draft and evidence is "missing" 3. Append every decision to ledger.jsonl with the answers, lane, audit flag and token estimate. Sample 1% of ignored events into queue/audit 4. Write → that prints jev cost, writing cost, jev share and escalation rate, and flags escalation above 6% Check the SDK attribute names against the installed version. Show me the plan first. Don't create files yet."
This post quotes another post: x.com/gippp69/status/2103394701063737504