Skales drops low-RAM desktop AI agent

Skales is interesting less because of raw agent novelty and more because it attacks the biggest adoption bottleneck: setup friction.

• –The real differentiator is packaging: EXE/DMG installers, ~300MB idle RAM, and local storage make it feel closer to a consumer desktop app than a research project
• –Ollama plus BYOK support gives it flexibility across privacy-sensitive local use and mainstream cloud workflows, which broadens its appeal beyond hobby demos
• –The feature surface is ambitious—Playwright automation, memory, voice, integrations, group chat, and a desktop buddy—which makes execution and reliability the real question
• –The BSL-1.1 license means this is source-available rather than true open source, so developers can study it but commercial adoption will hit licensing boundaries
• –If the UX holds up, Skales could carve out a lane for “AI agent for normal computers” instead of competing purely on benchmark-style agent capability