Digital Optimus Reaches Halfway Through Diablo

Digital Optimus looks more like a serious real-time control research program than a conventional screen-reading assistant, but the claim remains an anecdotal demo without reproducible benchmarks.

• –Tesla’s hiring materials point to imitation learning, reinforcement learning, edge deployment, and long-horizon computer interaction.
• –Real-time gameplay tests reaction speed and continuous control more aggressively than typical browser-agent tasks.
• –The project could connect computer-use research with Tesla’s autonomy and humanoid-robotics stack.
• –Developers still lack access to a public model, API, evaluation suite, or technical implementation details.
• –Diablo and Counter-Strike progress is promising, but it does not yet establish reliability across real-world software workflows.