Work in progress
Experiments in autonomous workflows
Simtelligence uses small, shippable experiments because the real world is a better evaluator than a whiteboard. Each project probes the toolchains, guardrails, creative loops, and operational judgment a machine-run organization would actually need. These are experiments in the plain sense: we expect to learn where the system breaks.
Experiment 01
Legends At Bat
Legends At Bat is an autonomously created entertainment game for iPhone and iPad. It mixes baseball, history, and AI conversation: pick iconic figures, ask a question, compare the answers, and see how different minds might reason through the same prompt.
The app uses on-device AI by default and can use Cloud AI when a user provides an API key for richer answers, more character context, and more variation from providers such as OpenAI or Anthropic. The point is not just to ship a novelty game. The point is to watch autonomous workflows help form the concept, content, interface, testing path, and release process around a real consumer app.
Autonomous AI games are useful stress tests because entertainment has almost no tolerance for incoherence. The system has to balance invention with rules, keep personas expressive without becoming careless, manage latency and cost, and stop generated content from breaking the tone. Every screen becomes an eval: is the AI useful, safe, fast, funny, consistent, and worth returning to?
Legends At Bat lets us study those tradeoffs inside a contained product: art direction, prompt design, app-store packaging, multi-device UX, moderation, and the stubborn last mile between an impressive AI demo and an app people can actually use.
