PinnedLuis Valencia·Oct 6State Machines, Stop Reasons, and Bounded ExecutionState Machines, Stop Reasons, and Bounded Execution
PinnedInPython in Plain EnglishbyLuis Valencia·Oct 5Why Tool Access Is Not AuthorizationThis is Article 4 in the Cogentrex Agent Harness series.A response icon1A response icon1
PinnedInPython in Plain EnglishbyLuis Valencia·Sep 29Tool Contracts: The Boundary Between Intent and ExecutionThis is Article 3 in the Cogentrex Agent Harness series.
PinnedInPython in Plain EnglishbyLuis Valencia·Sep 21Built an AI Agent Harness from Scratch, Here’s What Frameworks Abstract AwayA practical architecture for model loops, tools, state, budgets, policy, evidence, and observability, before reaching for LangGraph…A response icon1A response icon1
PinnedInPython in Plain EnglishbyLuis Valencia·Sep 29Building a Typed ReAct Loop Without a FrameworkBuilding a Typed ReAct Loop Without a Framework
InPython in Plain EnglishbyLuis Valencia·Oct 26Sandboxing Agent-Generated CodeThis is Article 8 in the Cogentrex Agent Harness series. Previous: Durable Runs, Replay, and Auditability. Articles 1, 2, 3, 4, 5, and 6…A response icon1A response icon1
InPython in Plain EnglishbyLuis Valencia·Oct 13Durable Runs, Replay, and AuditabilityThis is Article 7 in the Cogentrex Agent Harness series. Previous: Context Is Not Memory. Articles 1, 2, 3, 4, and 5 are also published.A response icon3A response icon3
Luis Valencia·Oct 9Context Is Not MemoryThis is Article 6 in the Cogentrex Agent Harness series. Previous: State Machines, Stop Reasons, and Bounded Execution. Articles 1, 2, 3, 4…
InPython in Plain EnglishbyLuis Valencia·Jul 29From Prompt to Production 01/41-Anatomy of a Production AI Agent: What You Actually Need to ShipThe 3 AM Refund That Changed EverythingA response icon1A response icon1
Luis Valencia·Jul 2Model Availability Is Now an AI Governance RiskThe Anthropic Fable 5 and Mythos 5 export-control episode is a reminder that enterprise AI teams need to monitor more than latency, price…A response icon1A response icon1