How to Run 4-8 Parallel Coding Agents with tmux and Markdown Specs
Original: Parallel coding agents with tmux and Markdown specs View original →
Parallel AI Agent Development in Practice
Manuel Schipper has shared a months-long practical experience running parallel coding agents using a lightweight setup: tmux, Markdown files, bash aliases, and six slash commands. The post earned over 100 upvotes on Hacker News and sparked significant discussion about AI-assisted development workflows.
Three Agent Roles
Each tmux window is assigned a specific role:
- Planner: Builds Markdown specs for new features or fixes
- Worker: Implements from a finished spec
- PM: Backlog grooming and idea management
Feature Design Files: The Core Mechanism
The foundation of the system is the Feature Design (FD) file — a structured Markdown document containing: the problem being solved, all solutions considered with pros and cons, the final solution with an implementation plan, files that need updating, and verification steps. Each FD is numbered and tracked through an 8-stage lifecycle from Planned to Complete.
Practical Limits
Beyond 8 concurrent agents, decision quality degrades. The /fd-init slash command bootstraps the full system into any repository. This workflow demonstrates how developers can leverage multiple AI agents in parallel without complex orchestration infrastructure — a valuable model as AI-assisted development becomes mainstream.
Related Articles
A Goldman Sachs analysis found that despite massive AI investment, the technology contributed negligibly to US GDP growth in 2025.
GitHub added another enterprise model option to Copilot. Microsoft AI’s in-house coding model MAI-Code-1-Flash is now generally available for Copilot Business and Enterprise, with admin policy controls and usage-based billing.
The policy fight is no longer just about model benchmarks. Axios reports that U.S. officials have revisited tools such as Entity List threats, security advisories, procurement pressure, and hosting liability rules as cheaper Chinese open-weight models gain enterprise traction.