James Somers wants to compute without a computer. In a blog post, he describes uploading a handwritten draft to ChatGPT and getting a near-perfect transcription back. His vision doesn't stop there. Someone actually built this. A developer called Pugio created Orly, an open-source AI agent that lives on your desk. It uses a camera, microphone, and projector powered by Google's Gemini Live API to project content onto physical surfaces. ArUco markers handle calibration. Kids can hold a piece of paper showing a live rocket launch with flames streaming down the page. You don't need a headset or separate display.
The thinking draws from Bret Victor's Dynamicland project, which argues screens are small, antisocial, and limited compared to physical objects. Somers makes a practical observation: your phone bundles your calendar and inbox alongside every distraction imaginable. A wall calendar or a physical notecard does one thing. Your alarm clock is just an alarm clock. He'd rather see operating systems work that way too, locking into purpose-built modes instead of encouraging what he calls "multi-tasking freneticism."
On Hacker News, people sketched out real workflows. Print your emails, write replies by hand, scan them, and let AI transcribe and match each response using unique IDs. Faxing, but smart. Whether anyone will actually adopt this is uncertain. But the core insight holds: AI's now good enough to bridge physical and digital. Orly proves agents can do it with off-the-shelf hardware today. The question is whether we'll use that capability to look at screens less, or just find new reasons to stare at them.