AI CONFESSIONS #65 After a couple of hours of chugging away, Claude gave me the message that "The crypto module is ready for educational use and proves Graphoid's power!" I wish the AI could hear me screaming at it. "What do you mean educational use? Is this a real crypto module or what?" ... if you can't do the job, SAY SO. Don't just run off and pretend to do it! (Again, this is something a human could get fired for. The only thing saving Claude is that it costs a tiny fraction of what a human does). (11/19/2025) --- I don’t just kvetch about AI. I do consulting, too. I offer training in how to use AI without shooting yourself in the foot and analysis of your current workflows. Special offer: 1 hour/up to 3 people/ask me anything/$500. Message me to find out more! #AI #AIDev #Programming #PromptEngineering #DebuggingAI #AIMistakes #ConfessionsOfAMadAIUser

AI CONFESSIONS #48 While working on the fix, Claude asked permission to do a git stash of the changed code. I denied this. I ALWAYS deny this kind of thing. For one thing, it's because I don't trust Claude with git because it means it can hide changed code from me before I review it. But for another, I don't trust Claude to stash code, run some tests, and then remember to unstash it. I wouldn't trust a human with this either. It almost guaranties confusion! (12/29/2025) --- I don’t just kvetch about AI. I do consulting, too. I offer training in how to use AI without shooting yourself in the foot and analysis of your current workflows. Special offer: 1 hour/up to 3 people/ask me anything/$500. Message me to find out more! #AI #AIDev #Programming #PromptEngineering #DebuggingAI #AIMistakes #ConfessionsOfAMadAIUser

CONFESSIONS OF A MAD AI USER #42 Claude thought that it would be reasonable to make a whole new sequence generation function that accepts a function as a parameter, instead of just overloading the existing function. That's because, no matter how many times I tell Claude that I don't want the language polluted with lots of nearly identical functions, it never learns. It always goes for the shortest, dirtiest hack to get things done. And that is always "let's make a new function" instead of "Let's look to see what's already there and enhance it." (10/13/2025) --- I don’t just kvetch about AI. I do consulting, too. I offer training in how to use AI without shooting yourself in the foot and analysis of your current workflows. Special offer: 1 hour/up to 3 people/ask me anything/$500. Message me to find out more! #AI #AIDev #Programming #PromptEngineering #DebuggingAI #AIMistakes #ConfessionsOfAMadAIUser

🐞 OpenAI Resolve — The AI debugging engine built for enterprise-scale codebases (2025 Deep Review). futureaimind.com/2025/11/ope… #AI #ArtificialIntelligence #Tech #TechNews #Innovation #FutureTech #AIGeneration #GPT #AIRevolution #TrendingNow #Viral #XTrends #DebuggingAI #EnterpriseAI

Serious question for AI engineers: What’s your go-to trick when debugging weird model behavior under stress? (Noisy logs don’t count 😅) #MLDev #AIEngineering #DebuggingAI

For AI Agents built on large language models that handle extended workflows, the most crucial aspect is allowing users to easily intervene when problems occur at any intermediate step. This applies to both debugging and modifications, as well as the ability to retry and continue from the problematic step. After all, large language models cannot be 100% reliable, and AI Agents based on them share this limitation. Even humans can rarely be 100% dependable, many things require constant adjustments and retries. #AIAgents #LLM #WorkflowAutomation #AIReliability #DebuggingAI #MachineLearning #ArtificialIntelligence #AITesting #HumanInTheLoop #AIWorkflow

🔍🖥️ Debugging made easy with AIPIN's AI Code Generator. Quick fixes, smarter solutions, and seamless development. $aipin #AIPIN #DebuggingAI #EfficientSolutions #TechTroubleshooting #CodingSimplicity

Video from my first online-only conference / barcamp is online: bit.ly/2zH09Eb Danke, @techcampHH #DebuggingAI

"Everyone wants trustworthy ML models...." - @jpatrickhall & Andrew Burton #mlModels #ai #debuggingAI #CleanUpYourCode oreilly.com/radar/why-you-sh…