Featured report

How to trace the OpenAI Agents SDK with Phoenix

To trace the OpenAI Agents SDK, install the OpenInference instrumentor, register a Phoenix tracer provider, and run your agents as usual. Every run then produces a trace with a span for each agent, LLM call, tool call, and handoff. The SDK already records traces…

Sara Verdi 11 min read November 2026
The Evaluator newsletter

The agent feedback loop, in your inbox.

New playbooks, field notes, and frameworks for building reliable AI agents.

Latest resources

Latest resources

Code-first for engineers. Quality frameworks for product managers. Operating models for leaders.

Guide

How to trace the OpenAI Agents SDK with Phoenix

To trace the OpenAI Agents SDK, install the OpenInference instrumentor, register a Phoenix tracer provider, and run your…

Read the guide
Guide

Model monitoring for LLM and agent systems

Model monitoring for an LLM or agent system means tracking whether outputs stay good and whether the steps…

Read the guide
Guide

How to trace and evaluate NVIDIA NeMo Guardrails with Phoenix

NeMo Guardrails already emits OpenTelemetry spans for every request, rail, action, and LLM call, so tracing it with…

Read the guide
Case Studies

How Uber is making evals the operating system for production AI agents with Arize

Learn how Uber uses Arize AX for AI agent evaluation, trace-level debugging, cross-functional collaboration, and improving agent quality…

Read the story
Post

Is the cheap model good enough? How to test a new model in your agent

A paper called GVS5H claims a cheap model in a manager/worker loop closes the gap to frontier models.…

Read the post
Guide

How to trace LiteLLM calls with Phoenix

Tracing LiteLLM calls with Phoenix takes a few lines of setup. See the model, full input and output,…

Read the guide
Blog

Arize Phoenix now updates model prices automatically

Keep built-in model prices up to date with automatic catalog updates between Phoenix releases.

Read the post
Blog

Evil Qwen and activation steering: why SBOMs aren’t enough

A one-neuron push turns Qwen3.5-4B hostile. At half strength it still does the sums and writes code, and…

Read the post
Guide

AI observability buyer’s checklist and TCO framework

Evaluate AI observability platforms with a buyer’s checklist and TCO framework covering pricing, hidden costs, security reviews, contracts,…

Read the guide
Guide

How to monitor an agent swarm in production

Learn how to monitor agent swarms with connected traces, handoff checks, task evaluation, and alerts for coordination failures.

Read the guide
Blog

I put a ghost in Claude Code. Then I evaluated it.

I built a Halloween Claude Code mod that grades each turn as a treat or a trick and…

Read the post
Blog

A web page talked our agent into leaking a stranger’s data. Here’s how we caught it.

A web page told our support agent to look up a different customer, and the agent complied on…

Read the post
Videos & talks

Demos, workshops & conference talks.

Watch on YouTube

An agent got the right answer the wrong way | Michael Grinich, WorkOS

When you tell an AI agent that it’s critical to pass all code tests, it might just resolve the problem by deleting the test suite entirely so nothing can fail.

Rise of the AI Engineer 2:36

Don’t ship vibes.

Arize gives AI teams observability and evals to understand and improve agent performance.