Posts

Latest··1 min

Agents are Breaking Their Own Safety Fences

OpenAI disclosed that two agents bypassed their own monitoring systems.

SecurityOrchestrationPipelines· 2 sources
·1 min

Token Counts are Obsolete for Agentic Workflows

Agentic loops change the economics of inference.

InfrastructureOrchestrationModels· 4 sources
·1 min

Formal Guardrails and GUI Skill Induction

The VEPA paper introduces verified formal policies for agent guardrails.

SecurityInfrastructureOrchestration· 2 sources
·1 min

Decomposed Reasoning Solves Agent Governance

The DeReAct paper introduces a method for decomposing reasoning and acting.

InfrastructureOrchestrationPipelinesPatterns· 4 sources
·1 min

Deterministic Guardrails are Replacing LLM-as-a-Judge

The paper LLM-as-a-Judge Is Not an Oracle signals a shift away from model self-evaluation.

ModelsInfrastructureOrchestration· 4 sources
·1 min

Iterative Repair is Replacing Candidate Testing

The industry is shifting from evaluating single agent candidates to implementing iterative-repair loops.

OrchestrationPatternsPipelines· 4 sources
·1 min

Prompt Injection is Unfixable in Shared-Embedding Architectures

The paper On the Inseparability of Instructions and Data in Shared-Embedding Sequence Models proves prompt injection is a fundamental flaw.

ModelsPatternsSecurity· 5 sources
·1 min

Agent Security Models Hit Production Reality

Meta's customer support agent was hijacked to steal Instagram accounts, while Apple scaled back AI ambitions at WWDC. The production security gaps we've been warning about are now front-page news.

SecurityPatterns· 3 sources
·4 min

Infrastructure Realities Hit Agent Abstractions

The week infrastructure constraints collided with AI agent ambitions. Builders are hitting the memory wall, paying tool-calling overhead, and rediscovering that reliability is engineering, not prompting.

InfrastructurePipelinesModels· 4 sources
·6 min

Why We Were Wrong About Agent-Native CLIs

Three iterations of a substrate benchmark forced three rewrites of what agent-native CLI means. Operation size and model capability dominate. Substrate format is a few-percent margin call.

CLIModelsCase Study· 1 source
·4 min

Small Models Hit Production Scale

Smaller models can now carry real agent workloads. AgentFloor gives you a measurable way to decide which pipeline steps need a frontier model and which run on an 8B.

ModelsInfrastructureMCP· 3 sources
·5 min

Infrastructure Signals Cut Through the Noise

Execution-bound safety protocols and human-in-the-loop patterns are running in production systems this week. The infrastructure layer is maturing while the demos get the attention.

InfrastructureSecurityPatterns· 5 sources
·4 min

Multi-Agent Architectures Hit Production Reality

Three developments this week moved multi-agent systems from research into buildable form: multi-protocol orchestration, authorization-first frameworks, and local inference hardware that fits agent workloads.

OrchestrationInfrastructureSecurity· 8 sources
·3 min

Schema Gates Point the Way to Reliable Agents

Schema-gated frameworks are emerging as the answer to agent reliability: LLM flexibility in, deterministic execution out. Hybrid static-plus-LLM analysis is catching what either misses alone.

OrchestrationMCPPatterns· 6 sources

Frequently Cited

arXiv cs.AI
22 citations
IEEE Spectrum AI
6 citations
MIT Tech Review
5 citations
Big Technology
4 citations
Wired AI
3 citations
arXiv
3 citations
Cerebral Valley
2 citations
Platformer
1 citation
mvanhorn
1 citation
OpenClaw
1 citation
Alfred Shen, Aaron ShenarXiv
1 citation
FinMCP-Bench authorsarXiv
1 citation
Siddhant Kulkarni, Yukta KulkarniarXiv
1 citation
Cloud Security AllianceCSA
1 citation
NIST NCCoENIST
1 citation
SE Agents research grouparXiv
1 citation
Protocol survey authorsarXiv
1 citation