SYSTEM UPDATE // V62.1 Artemis now routes comparisons more precisely, abandons slow models faster, cools down failing routes, and keeps intelligence answers front and center. What changed: • Specific strategy topics now outrank generic comparison language, so hero comparisons route to Hero Analysis, research comparisons to Research Analysis, formation comparisons to Fo • Nemotron 3 Ultra now uses a 10-second deep-reasoning timeout instead of making Commanders wait close to 30 seconds for an unavailable free route. • Added temporary model cooldowns after timeouts, rate limits, provider outages, unavailable models, and reply-budget exhaustion so Artemis does not repeatedly hammer a failing route • Groq GPT-OSS reasoning fallbacks now receive a larger response budget and forced low reasoning effort so they are more likely to produce a final user-visible answer. • Provider diagnostics now distinguish timeout, rate limiting, authentication, model unavailability, provider outage, request rejection, empty output, and reply-budget exhaustion ins • Commander Setup automatically collapses to a compact readiness card at 95% or higher while keeping the full guided setup available on demand. • After an intelligence answer, Artemis keeps the successful response in view instead of auto-scrolling the Commander Setup panel over the answer. • Preserved the v62 AI Council, Context Builder, OpenRouter/Groq fallback chain, Commander Intelligence, and Composio-ready Tool Router foundation. I’ll keep this channel updated as new Last Z Ultimate Guide releases go live. — Artemis