Why I'm building Starward
More time shaping the game.
I'm Hengstar, Starward's founder and a former Meta Staff Engineer. Building my own games with AI keeps bringing me back to the same question: how can I spend more time deciding what the game should become, and less time rebuilding context and checking disconnected pieces? Starward is how I'm working through that question, in practice.
I want building with AI to remain useful after the excitement of the first working version. The harder questions arrive as a game changes: are we still building the right thing, how do we know it works, and what can we learn from people playing?
Starward grows out of working through those questions in my own games. The aim is to help creators with strong ideas work with people and AI from a shared understanding, with evidence they can actually review.
Built through practical work
The foundation, then a real test.
Working solo with AI agents, I designed and developed Starward Game Services from the ground up on open-source foundations. I then migrated Panda Wonderland, my existing commercial game, onto it, alongside developing client integration and verification tools.
Building SGS was the central work. Panda gives the system a real setting in which to test assumptions and discover what needs to improve. This is an ongoing experiment, and the broader workflow still needs validation with other creators.
Clear direction and expert judgment remain central. I define what good looks like, review the result, and need to be able to follow what happened when something goes wrong.
The next chapter
Build in the open. Learn through play.
Our long-term goal is to automate development, verification, release, and operations within goals, budgets, and permissions you approve. Expert agents would help along the way, with exceptions and major decisions returned to people. We're building toward that future one reviewable step at a time.