Louis Arge

Arge’s personal site says he works at Gradients on technology intended to improve wellbeing and reduce extreme suffering. It describes the company’s work on pain therapeutics and lists Próspera and Teton.ai under previous work. His own projects also include a solar-system explorer, flashcards inside AI assistants, poker-learning tools, and an autocomplete tool for Obsidian.[4]

Helping separate agents coordinate

claude-peers runs a local broker that registers active Claude Code sessions and relays messages between them. Each session can discover peers by machine, directory, or repository, see a short description of their work, and send a message to a particular peer. A SQLite database holds broker state, and channel support pushes incoming messages into the receiving session. This gives independent sessions a way to ask who owns a file or share a finding without making the user copy information between terminals.[3]

Database work at the keyboard

dbdiff is a desktop PostgreSQL client for exploring, editing, and querying databases. Its README emphasizes keyboard-driven use and source installation with Node.js. The implementation combines an Electron desktop application with a Vite interface and Express service. It offers a more conventional developer tool alongside Arge’s agent experiments: a direct interface for inspecting and changing the data an application uses.[2]

Arge’s Skyline experiment explores social feeds organized around a user’s expressed preferences. His account describes embedding candidate Bluesky posts and filtering them according to the content someone wants more or less of, instead of ranking only by engagement. He built the client to examine whether a feed could preserve social connection while reducing material a person does not want to spend time reading. The post documents a personal experiment, not a controlled wellbeing study.[5]