Source note2 min read
Steerability via constraints: a substrate for scalable oversight of coding agents
Source
arXiv
arxiv.org/abs/2607.02389v1
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 proposes a system for managing coding agents that replicates methods used for decades in managing large human engineering teams, such as access control and strict coding conventions enforced by tooling. The system includes a controlled experiment where a reviewer inspects a Python codebase containing backdoors, demonstrating an increase in recall from 54.5% to 90.9% with the use of the proposed system.
What the Document Says
- The text discusses coding agents and human oversight.
- It mentions that unconstrained agents can pose security risks, reduce codebase scalability, and make human review costly.
- The author argues that methods used for decades to manage large human engineering teams can be applied to coding agents, specifically access control, network policies, and strict coding conventions enforced by tooling.
- The text mentions a start-to-end system based on this principle.
Key Concepts Mentioned
- coding agents
- human oversight
- security risks
- codebase scalability
- access control
- network policies
- strict coding conventions
- tooling
- controlled experiment
- Python codebase
- backdoors
- recall
- substrate-level oversight
Claims Made by the Document
- Unconstrained agents can pose security risks, reduce codebase scalability, and make human review costly.
- Methods used for decades to manage large human engineering teams can be applied to coding agents.
Evidence Used
- The text states that unconstrained agents can introduce security risks, erode codebase scalability, and make human review increasingly costly.
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.
