IDE-centric, not lifecycle-aware. Cursor is an excellent code editor with AI, but it does not provide a workflow pipeline, merge conflict resolution, or provenance. Its "multi-agent" mode is best-of-N (same prompt x3), not role differentiation, so it is complementary with hashd rather than a direct replacement.

Google Antigravity is an agent-first IDE with parallel dispatch and Gemini 3, and it is genuinely impressive for greenfield parallel development. But its "multi-agent" story is mostly marketing for parallel task dispatch rather than coordinated review, and the conflict strategy is often "keep agents in separate folders." The D: drive deletion incident, five critical security vulnerabilities within 24 hours of launch, 1GB+ idle memory use, and cloud tethering are meaningful concerns despite aggressive free pricing.

GitHub Copilot has 150M+ users and best-in-class inline completions. Copilot Workspace is a step toward orchestration but remains single-task, and their own SDK issue tracker says it "lacks primitives for autonomous multi-agent workflow orchestration." hashd orchestrates what Copilot produces.

OpenAI Codex is a great single-task execution agent, and hashd uses Codex as its default implementation agent. Codex is excellent at "implement this specific commit." hashd adds everything around it: orchestration, review loops, merge conflict strategy, and provenance.

Entire.io has serious backing ($60M seed, ex-GitHub CEO) and asks the right question: "Git was built for human-to-human. What does the AI era need?" But Entire is primarily an observability layer that records the archaeology of AI-generated code, while hashd orchestrates how that code is created. The products are complementary.

Devin (Cognition Labs) is fully autonomous and async-only, but at $500/month it is expensive and offers a binary HITL model rather than a configurable autonomy spectrum. It lacks a per-commit review loop and oscillation detection, and it is not open source.

Factory.ai is the closest competitor in orchestration depth, with HyperCode codebase understanding and custom Droids. But it has no explicit FSM model, is not open source, and important workflow details remain unclear.