Blog

From the OLAV team

Release announcements, architecture deep-dives, and notes on building AI-native infrastructure tooling.

September 19, 2026 · OLAV Team decision-call jev netdevops agent infrastructure performance

Stop Making LLMs Emit JSON: How Decision Calls and Jev Are Shedding Autoregressive Overhead in NetDevOps

In high-availability infrastructure, waiting 2.4 seconds for an LLM to generate JSON during micro-bursts causes cascading failure. Here is how single-forward-pass Decision Calls and Jev deliver calibrated probabilistic decisions in 70ms.

September 2, 2026 · OLAV Team skills pi-agent netops portability open-source agent-runtime

We Said Any Agent Could Run These Skills. Here's One That Isn't Claude Code.

olav-skills and olav-presales-skills claim no framework binding — just a SKILL.md and a stdin-JSON script. We installed pi-agent, a different vendor's runtime with its own model routing, on a bare machine and let it drive the skills unsupervised. It found a real bug, fixed it, and wrote a report we could verify against the database.

August 19, 2026 · OLAV Team netops skills collector claude-code enterprise workflow

Your Agent Subscription Already Runs Network Analysis — It Just Cannot Reach Your Network

Two open repos put a real network model in front of the agent platform your company already pays for: olav-collector on the jump host, olav-skills on the workstation, and nothing but a directory of text between them.

August 9, 2026 · OLAV Team security AI agent network automation SSH MCP

SSH MCP Gives AI a Universal Key — OLAV Gives AI Constrained Capability

Why raw SSH access for AI agents creates hallucination death spirals, and how structured tools with deterministic verification change the equation.

July 25, 2026 · OLAV Team Architecture NetOps AI Agent Log Analysis Root Cause Analysis

OLAV vs. Log-Based AI: Why Network Operations Needs More Than a Smarter grep

Most 'AI for operations' tools are just putting an LLM on top of log search. OLAV takes the opposite approach: model the network engineer's entire workflow — collection, parsing, layered diagnosis, snapshot comparison, simulation, and change planning — as a team of AI specialists. Here's why that matters and how it works.

July 22, 2026 · OLAV Team Continual Learning Agent Architecture Local LLM Fine-Tuning Self-Improvement

One Agent, Three Ways to Learn: How OLAV Builds a Self-Improving Agent

Learning isn't one thing. OLAV's agent improves across three layers at three speeds — memory that learns every run, weights it can consolidate into with a benchmark gate, and a harness that tunes its own knobs — unified by one loop and one governance contract so they reinforce instead of fight. Here's the story of how we built it, including the times the system correctly threw its own work away.

July 21, 2026 · OLAV Team Architecture Small Models Agents Local LLM NetOps

Building Agent Systems on Small Local Models

Inside an enterprise network you often can't send data to a hosted model, and a local frontier model is too expensive to run. A ~30B model is the practical sweet spot — but it hallucinates, its knowledge is thin, and its usable context is short. Here's how to build a real agent on top of it anyway.

July 21, 2026 · OLAV Team Quick Start NetOps Local LLM Nornir Batfish

OLAV Quick Start: Your Network In, Answers Out — on a Local Model

A hands-on 15-minute quick start. Install OLAV, get your fleet in two ways (live SSH collection with Nornir, or an offline backup import), take a fresh snapshot, then ask in plain English, draw the topology, draft a validated change, and run a morning audit — all on a local ~30B model, nothing leaving your network.

April 9, 2026 · OLAV Team release devops infra netops architecture security