Changelog

What's new at Devence Lab

Articles we have published, positions we have opened, and where it started, newest first.

September 18, 2026
Why Static Interpretability Fails on Multi-Step Agentic Decision Chains
- ArticleA technical account of the mismatch between single-forward-pass interpretability methods and the failure modes of agents that plan, act, observe, and revise over many steps - grounded in published circuit-analysis literature, a controlled failure-attribution benchmark, and emerging graph-based tracing methods.
September 17, 2026
Open positions
September 16, 2026
The Agent Sandbox: A Reference Architecture for Isolating Autonomous AI Systems
- ArticleAn industry-neutral engineering reference on containing agents that browse, execute code, and act on the open internet - synthesized from disclosed CVEs, published isolation-technology internals, OWASP's agentic risk taxonomy, and the operating architectures of production agent-sandbox vendors.
September 14, 2026
Runtime Monitors for Autonomous Systems: Detecting Drift and Misbehavior After Deployment
- ArticleA methodological deep dive into why evaluation is not oversight, how the runtime-verification and machine-learning-monitoring traditions converge on autonomous language-model agents, and what the published, quantitative record says about how well post-deployment monitors actually work.
September 12, 2026
Case Study: Using Interpretability to Catch a Specific Failure Mode Before Deployment
- ArticleA technical breakdown of three verified, published research programs that used mechanistic interpretability, linear probes, sparse autoencoders, and residual-stream analysis, to detect deception, hidden reward-hacking objectives, and strategic underperformance in language models before those models reached real users.
September 10, 2026
Sparse Autoencoders: What They Reveal, and the Accuracy Tradeoffs Nobody Advertises
- ArticleA methodological deep dive into sparse dictionary learning for neural network interpretability, how sparse autoencoders (SAEs) work, what they have genuinely revealed, and the specific, measured costs that get quietly left out of the demo.
September 8, 2026
Attribution Graphs: A Technical Walkthrough of Circuit Tracing, and Where It Breaks
- ArticleA methodological deep dive into how Anthropic's interpretability team traces the internal computation of large language models, what the resulting "attribution graphs" actually show, and the specific, documented ways the method fails.
September 6, 2026
Chain-of-Thought Monitoring: A Fragile Window Into Model Cognition
- ArticleA technical deep dive for engineers and researchers building oversight into agentic systems
May 2, 2026
devencelab.com launches
- NewThe first version of the Devence Lab website.

What's new at Devence Lab

Why Static Interpretability Fails on Multi-Step Agentic Decision Chains

Open positions

The Agent Sandbox: A Reference Architecture for Isolating Autonomous AI Systems

Runtime Monitors for Autonomous Systems: Detecting Drift and Misbehavior After Deployment

Case Study: Using Interpretability to Catch a Specific Failure Mode Before Deployment

Sparse Autoencoders: What They Reveal, and the Accuracy Tradeoffs Nobody Advertises

Attribution Graphs: A Technical Walkthrough of Circuit Tracing, and Where It Breaks

Chain-of-Thought Monitoring: A Fragile Window Into Model Cognition

devencelab.com launches