---
title: "Discussion on \"The Behaviour Was the Bug: Observability for Systems Where Stack Traces Don't Help\" | Hashnode"
description: "Discussion on \"The Behaviour Was the Bug: Observability for Systems Where Stack Traces Don't Help\". Every code path executed correctly. The HTTP response was 200. The answer was wrong. Classical observability has nothing useful to say about that failure. This article is about the instrumentation discipline that does."
url: "https://hashnode.com/posts/llm-observability-when-stack-traces-fail/6a58494b28416a9798790d34"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [The Behaviour Was the Bug: Observability for Systems Where Stack Traces Don't Help](https://pragmaticstack.in/llm-observability-when-stack-traces-fail?utm_source=hashnode&utm_medium=feed)

Every code path executed correctly. The HTTP response was 200. The answer was wrong. Classical observability has nothing useful to say about that failure. This article is about the instrumentation discipline that does.

Jul 16 · 15 min read · pragmaticstack.in  
Read the article

-   [#observability](https://hashnode.com/tag/observability)
-   [#llm](https://hashnode.com/tag/llm)
-   [#ai](https://hashnode.com/tag/ai)
-   [#software-architecture](https://hashnode.com/tag/software-architecture)
-   [#devops](https://hashnode.com/tag/devops)

## Written by

[Alen Joy](https://hashnode.com/@alenj)

@alenj · Bangalore, India

  
[View profile](https://hashnode.com/@alenj)

I help enterprises ship AI that doesn’t embarrass them in production.

AI Architect. I design the boring control plane plumbing that keeps your impressive demos from quietly setting themselves on fire in production.

## Responses 5

Markdown supported · ⌘↵ to post

[Mateo Ruiz](https://hashnode.com/@mateoruiz)

[Jul 16](https://hashnode.com/posts/llm-observability-when-stack-traces-fail/6a58494b28416a9798790d34/comment/6a58d2a2874ccfbb8432706a)

This really resonated, especially the point that the giveaway isn't using AI it's not editing what AI gives you.

We've noticed the same pattern at IT Path Solutions when reviewing AI-generated MVPs. The code or UI is often 80–90% there, but the last 10% refining UX, questioning defaults, improving copy, and removing template artifacts is what makes a product feel intentional instead of generic.

AI has made building fast. Judgment, iteration, and attention to detail are what make something memorable. That's becoming a much stronger differentiator than simply shipping first.

[Alen Joy](https://hashnode.com/@alenj)

[Jul 21](https://hashnode.com/posts/llm-observability-when-stack-traces-fail/6a58494b28416a9798790d34/comment/6a58d2a2874ccfbb8432706a#reply-6a5f7aeac7155c060fe70e12)

I completely agree. AI has dramatically reduced the cost of getting to a first draft, but it hasn't reduced the value of judgment. That final refinement, questioning assumptions, improving UX, tightening copy, and removing the "AI fingerprints" is where products start to feel crafted rather than generated. I think that's where teams will increasingly differentiate themselves.

[Mayank Bansal](https://hashnode.com/@monk1)

[Jul 16](https://hashnode.com/posts/llm-observability-when-stack-traces-fail/6a58494b28416a9798790d34/comment/6a58596f349b1e2230c12dbf)

I think we're starting to see two separate infrastructure layers emerge. One is observability capturing traces, prompts, retrieval context, and behaviour after the fact. The other is runtime enforcement policies, completion validation, and execution checks before an action happens. Neither replaces the other, but together they make production agents much easier to trust.

[Alen Joy](https://hashnode.com/@alenj)

[Jul 21](https://hashnode.com/posts/llm-observability-when-stack-traces-fail/6a58494b28416a9798790d34/comment/6a58596f349b1e2230c12dbf#reply-6a5f7a957d89f1c40b5e111a)

Well put. I see them as complementary control planes. Observability gives you the evidence to improve the system, while runtime enforcement provides the guardrails that keep failures from propagating in production. As agentic systems become more capable, that separation between "observe" and "govern" is becoming a useful architectural pattern.
