Blog

Field notes on agent behavior.

Teardowns of real failures, what the traces showed, and what we learned building observability for agents in production.

GuideFeatured

AI Agent Failure Modes in Production: The Complete Taxonomy (2026)

A working taxonomy of the AI agent failure modes that traces and evals miss: false success, loops, silent drift, intent mismatch, and cohort gaps, with what each looks like in a real session.

Alexandre Ayoub · Jul 18, 2026 · 6 min

se_19f2c4

4 turns · 14.3s

The speaker arrived cracked. I'd like a refund.

process_refund402

PAYMENT_PROVIDER_DECLINED

Your refund of $49.00 has been processed.flagged

Guide

The 9 Best AI Agent Observability Tools in 2026

Agent observability has split into three layers teams routinely confuse: tracing, evals, and behavioral observability. Here are the nine tools we see most in production stacks, what each is actually for, and where each one stops.

Jul 3, 2026 · 11 min

Architecture

What is behavioral observability?

Behavioral observability is the practice of detecting how an AI agent behaves across sessions and users, not just whether each LLM call succeeded. Here is the definition, the signals, and how it differs from execution observability.

Apr 25, 2026 · 9 min

Engineering

How to detect agent drift in production

Agent drift is the failure mode every AI team talks about and nobody measures. Here is how to detect it, which signals matter, and how structured memory stops it.

Apr 10, 2026 · 10 min

Engineering

How to integrate Flowlines in 5 minutes

Add behavioral observability and structured memory to any Python AI agent. Install the SDK, init before your LLM client, wrap calls in context, and retrieve memory. Works with OpenAI, Anthropic, and any agent framework.

Mar 15, 2026 · 5 min

Opinion

Flowlines vs. Mem0: why memory needs observability

Comparing Flowlines and Mem0 for AI agent memory. Mem0 is a write API. Flowlines builds memory from observed behavior. Here is when each one makes sense.

Mar 8, 2026 · 7 min

Architecture

Intent engineering requires memory

Intent detection treats every message as isolated. But real users have evolving goals. Without AI agent memory, intent resets every turn. Structured memory changes that.

Feb 27, 2026 · 6 min

Architecture

LLMs are stateless. Systems shouldn't be.

LLMs are stateless by design. Every API call starts from zero. But production AI agents need continuity and structured memory. That gap is an infrastructure problem.

Feb 23, 2026 · 7 min

Engineering

The silent failure problem in AI agents

Most AI failures don't crash. The agent returns a plausible answer and moves on. Without observability and structured memory, these silent failures repeat forever.

Feb 2, 2026 · 8 min

Opinion

Why AI agents don't learn in production

Production AI agents look intelligent in demos. But they don't get better over time. Every session starts from zero. The missing piece is AI memory infrastructure.

Jan 25, 2026 · 9 min

Get the field notes

Teardowns and findings, roughly twice a month. No digests, no product news.

Subscribe via RSS