AI Safety / World Simulations
A speculative build track around simulated environments, agent behavior, and safety-relevant failure modes. Honest, early, and grounded in learning by building.
hi, i'm ben(jamin) johns
leading ai deployments at scale
now
building toward work that still feels worthwhile in ten years.
i am focused on reinforcement-learning implementation, applied ai systems, and the habits of clear technical execution. the thread underneath is simple: make useful things, understand them deeply, and write down what becomes clearer.
recents
Claude / Gong / Rocketlane Hackathon
A generic customer-signal workflow: meeting transcripts and account context become Slack approval loops and project tasks with traceable human review.
AI Deployment Is a Product Discipline
A short note on why useful AI work depends on implementation, adoption, evaluation, and operational patience.
World Simulations as a Safety-Relevant Building Direction
A speculative but grounded note on environments, agents, and learning by building small simulations.