Matt Van Horn builds open-source tools for AI agents, including last30days, CLI Printing Press and agentcookie. He co-founded June, the smart-oven company acquired by Weber. His earlier career included Zimride, Digg and Path.[1][2][6]
Start with Last30Days AI Research Skill →
Education
Links
GitHub 
X LinkedIn Printing Press Career
Van Horn graduated from the University of Arizona in 2006 with a business degree in marketing and entrepreneurship. His early career included Zimride, Digg and Path before he co-founded June with Nikhil Bhogal in 2013. The company built a connected countertop oven. Weber acquired June in January 2021 and appointed Van Horn president of the June business. His current public work centers on open-source software for agents; the acquisition announcement does not establish that he still holds that executive role.[2][3][6][1]
Contributions
Researching recent discussion with last30days
The last30days skill searches recent material across public discussion and information sources, then asks an agent to synthesize the findings. Its repository describes sources including Reddit, Hacker News, X and YouTube, with access depending on the available configuration. This helps an agent find current examples and conversations beyond its training data. Engagement is one input to the tool's ranking, not proof that a popular claim is accurate, so readers still need to inspect the underlying sources.[4]
Giving agents command-line tools
CLI Printing Press generates command-line interfaces for APIs and websites. Its workflow researches an integration, builds the tool, checks it and packages related agent instructions. The project emphasizes structured commands, local SQLite copies and offline search so an agent can query useful data without repeatedly reconstructing an integration. The generated interface still needs testing against the service it connects to; a generated command is not evidence that every API behavior has been verified.[5]
Keeping browser sessions available
agentcookie addresses a practical problem for people running an agent on a separate machine: the agent's browser often lacks the user's existing authenticated sessions. The current repository describes transferring permitted Chrome cookies from a Mac to a Linux agent machine over Tailscale, then injecting them into the running browser. It also documents a limit: device-bound Google sessions cannot be transferred this way and require a sign-in on the destination.[7]
Tools, writing & talks
Tools
Start here
Last30Days AI Research Skillgithub.comAn agent skill that searches recent discussion and source material, then assembles a research brief.
Read the README and try one focused research question in a supported agent. You will need a coding-agent environment; inspect its linked sources to judge whether the resulting brief supports its conclusions.
Project creator[1][4]Original source- agentcookiegithub.com
Synchronizes permitted browser sessions to an agent machine, with documented limits for device-bound authentication.
Project creator[1][7]Original source - CLI Printing Pressgithub.com
A workflow for generating command-line interfaces and related agent integrations for APIs and websites.
Project creator[1][5]Original source
Writing, research & talks
A viral article by Matt Van Horn breaking down the 15 most common AI agent loops people are actually running, complete with copyable commands. It's the sequel to a 3.6M-view piece on what loops are, now focused on practical implementation.
Catalog workOriginal sourceAn article/post arguing that AI agents should store learned lessons as pull requests to shared skill files rather than private memory files. The core insight is that private memory accumulates noise and makes agents dumber over time, while code-reviewed skill files keep agents sharp and auditable. The author advocates limiting memory to 6 files as a forcing function for quality.
Catalog workOriginal source