---
title: "Discussion on \"The Founder Who Tested His Own Production App With AI — And Regretted His Confidence \" | Hashnode"
description: "Discussion on \"The Founder Who Tested His Own Production App With AI — And Regretted His Confidence \". I Thought I Knew My App. Two AIs Disagreed.\nHow 23 minutes of Passmark testing taught me more about my own production app than two years of building it.\n\nThere is a particular kind of confidence that"
url: "https://hashnode.com/posts/the-founder-who-tested-his-own-production-app-with-ai-and-regretted-his-confidence/6a0140b1e3eebc2e20f040bd"
---

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

Discussion

[Priyanshu Anand](https://hashnode.com/@priyanshuworks)

May 11

# [The Founder Who Tested His Own Production App With AI — And Regretted His Confidence](https://priyanshudev.hashnode.dev/the-founder-who-tested-his-own-production-app-with-ai-and-regretted-his-confidence?utm_source=hashnode&utm_medium=feed)

I Thought I Knew My App. Two AIs Disagreed. How 23 minutes of Passmark testing taught me more about my own production app than two years of building it. There is a particular kind of confidence that

priyanshudev.hashnode.dev 16 min read

[#breakingappshackathon](https://hashnode.com/tag/breakingappshackathon)

## Responses 2

Markdown supported · ⌘↵ to post

[Vishal](https://hashnode.com/@vk2037)

[May 13](https://hashnode.com/posts/the-founder-who-tested-his-own-production-app-with-ai-and-regretted-his-confidence/6a0140b1e3eebc2e20f040bd/comment/6a04432852a9c969d62d399c)

How many plain-English assertions did you write total, and what was the pass/fail ratio on first run? For the state- management failures - were these triggered by a specific sequence of user actions, or did they appear randomly? Trying to understand if the AI is doing path-coverage testing or just linear flow execution.

[Priyanshu Anand](https://hashnode.com/@priyanshuworks)

[May 14](https://hashnode.com/posts/the-founder-who-tested-his-own-production-app-with-ai-and-regretted-his-confidence/6a0140b1e3eebc2e20f040bd/comment/6a04432852a9c969d62d399c#reply-6a052e97a0c154027793f6ac)

I actually wrote 52 assertions in total, but featured the top 10 in the article for clarity. On that first run, I hit a 3/7 pass-fail ratio for those featured tests. The state-management issues were triggered by specific sequences the AI discovered through path exploration; it wasn't just a linear script, but an agent finding edge cases by interacting with the UI in ways I hadn't anticipated