Continuous AI: Important, not urgent
Why fixed-cost capacity changes what we're willing to let agents investigate — and why they should bring us proof instead of a bigger backlog.
LunaRoute blog
Field notes from building LunaRoute: how fixed-price inference works, what it takes to run it, and what we learn shipping it.
Introducing LunaRoute Cybersecurity and Cyberdefense
Two new LunaRoute programs for authorized security work: Cybersecurity for vulnerability research and testing, Cyberdefense for threat analysis, detections and response.
Meta Muse Spark Open is Now Available on LunaRoute
Meta Muse Spark Open is now on LunaRoute — a capable US open-weight model for tool use, coding, and long-horizon agents, on US servers with Zero Data Retention and fixed-cost inference.
The Model Is Not the Product
Same model, wildly different experience: why quantization, caching, and scheduling — not the weights — decide what your app actually gets.
Introducing LunaRoute
Today we're launching LunaRoute: private AI capacity at a fixed cost, so you can keep experimenting without watching a token meter.