Dev

dev-3.0 is a sharp, opinionated tool for a narrow but underserved niche: solo developers running a fleet of AI coding agents in parallel. Its core insight—that context-switching and branch conflicts are the real bottleneck—is well-executed via per-task git worktrees and live terminal previews. The Kanban board is not just eye candy; it's functional, letting you drag tasks, label them, and search across them. The 'one task, many attempts' feature is genuinely useful for comparing agent outputs. Where it shines is highly parallel, multi-agent workflows. If you routinely launch Claude Code for one feature, Codex for another, and Gemini for a refactor—all on different branches—dev-3.0 keeps each sandboxed and visible. Column agents and auto-review add automation that saves real time. However, it's not for everyone. Beginners will find the terminal dependency (tmux, Ghostty) intimidating. There's no collaboration: no shared boards, no user roles. It's also alpha software—expect bugs and missing polish. Pricing hasn't been announced; it's free now, but the model might change. Compared to tools like Cursor or Windsurf that embed agents in an IDE, dev-3.0 is the opposite: it stays out of the editor and orchestrates agents purely through the terminal. If you prefer visual agent chat (like ChatGPT desktop), look elsewhere. If you're already comfortable with tmux and git worktrees, dev-3.0 could make you significantly faster. In practice, we'd reach for this when running 3+ agents in parallel on the same repo, or when comparing agent outputs side-by-side. For single-agent use or team col lab, it's overkill. The open-source, local-first approach is a plus for privacy-conscious users. Keep an eye on GitHub for maturity.