Steerability via constraints: a substrate for scalable oversight of coding agents

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Thomas Winninger
• Venue: arXiv: cs.AI
• Published: 2026-07-02
• Collected: 2026-07-03
• URL: http://arxiv.org/abs/2607.02389v1
• PDF: https://arxiv.org/pdf/2607.02389v1
• DOI: UNKNOWN

Plain-Language Brief

The paper suggests a method to manage coding agents, which can handle complex tasks but require human oversight due to security risks and scalability issues. The author proposes using the same methods used for managing large engineering teams (like access control, network policies, and strict coding conventions enforced by tooling) to oversee these agents. They present a system based on this principle and demonstrate its effectiveness through an experiment involving a reviewer inspecting a Python codebase containing backdoors.

What the Document Says

• Coding agents are capable but have security risks, erode scalability of codebases, and make human review costly.
• The author proposes using methods for managing large engineering teams to oversee coding agents.
• These methods include access control, network policies, strict coding conventions enforced by tooling.
• The proposed method is claimed to be cheaper than recent agentic scaffolding.
• A controlled experiment is conducted with a reviewer inspecting a Python codebase containing 11 backdoors.
• Recall improves from 54.5% (unconstrained, no tools) to 90.9% (constrained substrate plus a ~200-LoC docs CLI).

Key Concepts Mentioned

• coding agents
• human oversight
• security risks
• codebase scalability
• access control
• network policies
• strict coding conventions
• tooling
• substrate-level oversight
• Python
• backdoors
• recall

Claims Made by the Document

• The proposed method is cheaper than recent agentic scaffolding.
• The substrate-level oversight gains are largest where the language gives the fewest guarantees by default.

Evidence Used

• The author presents a system based on the proposed principle and demonstrates its effectiveness through an experiment.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Coding agents are capable; human oversight is the bottleneck. Unconstrained agents introduce security risks, erode codebase scalability, and make human review increasingly costly. We argue that the same methods used for decades to manage large human engineering teams: access control, network policies, strict coding conventions enforced by tooling; transfer directly to coding agents, and are cheaper (in token) than recent agentic scaffolding. We sketch a start-to-end system on this principle, and report a controlled experiment in scalable oversight: a small reviewer (Gemma 4 e4b) inspects a Python codebase containing 11 inserted backdoors. Recall rises from 54.5% (unconstrained, no tools) to 90.9% (constrained substrate plus a ~200-LoC docs CLI), with substrate and tools contributing independently. We choose Python deliberately: substrate-level oversight gains are largest where the language gives the fewest guarantees by default; the principles extend to languages like Rust.