The Nen Journal
Thoughts, engineering deep dives, and product updates on the frontier of desktop automation.
Prompt Pack Refinement: Distilling UI Agent Failures into Reliable Computer Use Agents
Nen turns existing desktop applications into infrastructure agents can reliably use. Today we're sharing **prompt pack refinement**, an inference-time technique that turns agent failures on recurring UI workflows into compact, reusable instructions that improve reliability withou...
How we used Nen to further AI research
Two weeks ago, the CRUX (Collaborative Research for Updating AI eXpectations) project ([link](https://cruxevals.com/)) led by researchers from Princeton University ran their inaugural experiment in which an AI agent built and published an app to the iOS App Store end-to-end, with...
Introducing the Nen Computer Use Desktops — Cloud Windows Desktops for Computer Use Agents
Most teams building computer use agents already own the agent loop and call out to a sandbox for execution. OpenAI made that pattern official this week with native sandbox support in the Agents SDK. Every sandbox provider on the launch list runs Linux. If parts of your workflows ...
What the Standout YC W26 tells us about Windows
YC Winter 2026 is officially in the books, and the data is staggering. According to Garry Tan, 14 companies hit the $1M ARR milestone before Demo Day and 10 are raising at over $100M valuations. Rebel Fund’s analysis was even more direct: **"No previous batch has come close to th...
How Computer Use Agents Speed Up Time-to-Value in Healthcare
If you're building AI agents for healthcare, you already know the integration nightmare. EHRs. Payer portals. Prior auth queues. Desktop software that runs the daily operations of clinics, hospitals, and billing teams — and was never designed with automation in mind. Here's the t...
Computer Use Infrastructure for Developers
After months of building in stealth, I'm excited to officially introduce Nen to the world. We started with a problem I kept running into as a developer: AI agents that talked a good game but couldn't actually operate production desktop applications. Everyone was building copilots...