Stab - Agent-native QEC simulation toolkit

• The description states that Stab is an agent-native toolkit for quantum error-correction (QEC) research.
• It is written in safe Rust and designed to be modifiable by researchers and their AI agents.
• Its first milestone is a drop-in replacement for Stim v1.16.0, implementing the same CLI commands, input/output formats, and result streaming.
• Circuit generator produces byte-identical output to Stim for supported arguments, including comments.
• The project uses Rust with zero unsafe code, enforced by linting and compiler checks.
• It is built using AI tools (Codex, GPT-5.6, GPT-5.5) and structured around autonomous agent workflows.

Note

No evidence of actual product usage or customer feedback beyond the author’s own development process.

• The description positions Stab as a “safe-Rust codebase” that allows researchers to safely modify and extend QEC simulation tools.
• It is described as an agent-native toolkit, implying integration with AI coding agents.
• The author frames it as a response to limitations in Stim — a widely used simulator — which lacks support for some advanced features like non-Clifford gates.
• The long-term goal is to build composable Rust components for QEC tooling, suggesting a shift from a single-purpose tool toward modular infrastructure.

Inference This evolution suggests a move from solving immediate compatibility issues to building foundational infrastructure. However, no evidence of prior positioning or market feedback exists.

• The target customer is described as quantum-computing researchers working on QEC.
• These users are said to currently rely on Stim but want capabilities beyond its current scope (e.g., non-Clifford gates).
• The author notes that modifying Stim requires expertise in systems code, which many QEC researchers lack.
• The tool is intended for use alongside AI coding agents.

Note

No evidence of specific customers or user groups beyond the author’s own research context.

• Not evidenced.

The description does not mention any pricing model, monetization strategy, or business model.

• Built with Codex and GPT-5.6 (and earlier versions of GPT-5), indicating reliance on AI-assisted development.
• Uses safe Rust with zero unsafe code, enforced by linting.
• Implements a structured workflow involving:
  • Explicit plans for each milestone
  • Repository-level instructions in AGENTS.md
  • Goal files to define completion criteria
• Includes custom review skills:
  • Full-code-review skill examining nine lanes (code quality, compatibility, CLI formats, performance, etc.)
  • Milestone-audit skill reviewing acceptance criteria and loopholes
• Performance is described as mixed — some paths outperform Stim, others need more work.
• SIMD kernels are isolated behind safe abstractions.

Inference The approach shows a deliberate attempt to manage complexity through AI agents and structured code review, but no evidence of production deployment or scalability beyond the author’s own use case.

• Not evidenced.

There is no mention of users, customers, revenue, adoption, or any form of traction beyond the author’s development efforts.

• The project is positioned as a potential alternative or extension to Stim, which is described as the field’s fast, battle-tested simulator.
• Stim supports workflows researchers depend on today, but some researchers want capabilities beyond its current scope.
• Stab aims to bridge this gap by enabling easier extension via AI agents and safe Rust.

Inference The competitive landscape includes Stim and potentially other QEC simulation tools. However, no evidence of market presence or competitive positioning is provided.

• Heavy reliance on AI coding agents without independent validation of correctness or performance.
• Limited human oversight despite large autonomous agent sessions (e.g., days-long runs).
• No clear evidence of real-world usage or feedback from QEC researchers.
• The author states that performance is still being optimized, suggesting early-stage maturity.
• Dependency on AI tools like Codex and GPT-5.6 raises concerns about alignment and reliability.

Inference The project may be in a very early stage with significant uncertainty around correctness, scalability, and real-world utility.

1. What specific QEC research workflows or use cases does Stab aim to support beyond Stim?
2. How do you validate that the AI agents are producing correct results, especially in performance-critical areas?
3. Have you tested Stab with actual QEC researchers or institutions? If so, what was their feedback?
4. What is the plan for expanding beyond non-Clifford gates and toward reusable Rust components?
5. Are there any known limitations or trade-offs between safety, performance, and extensibility in Stab?