🤖
AI for Developers
Using AI assistants without shipping their mistakes
AI for Developers
When Not to Reach for an AI Assistant
AI tools are powerful, but some situations call for restraint. Knowing when to switch them off is part of using them well.
9 min read
AI for Developers
Rubber-Duck Debugging, With or Without the AI
Explaining your problem out loud often reveals the answer before anyone responds. An AI assistant can be a very patient duck.
9 min read
AI for Developers
Why You Still Have to Review AI-Generated Code
AI assistants are genuinely useful and confidently wrong in ways that are easy to miss. Treating their output as a draft, not an answer, is essential.
9 min read
AI for Developers
Prompting an AI Assistant Well: Clear Context Beats Clever Tricks
Getting good results from an AI coding assistant is less about secret phrases and more about the same clarity you’d give a new colleague.
9 min read
AI for Developers
Prompt Injection: When Your Data Becomes an Instruction
If your application feeds fetched text to a model, that text can give the model orders. Why filtering does not fix it, and the boundary that actually contains it.
9 min read
AI for Developers
When an AI Assistant Invents a Function That Does Not Exist
Generated code that calls a plausible, non-existent API is the failure mode most likely to reach your main branch. Why it happens, why it survives review, and the checks that catch it.
9 min read
AI for Developers
Saturation in Neural Networks: Where Gradients Go to Die
A saturated activation function stops learning, not because the model is wrong but because its gradient has gone nearly to zero. The specific mechanism, and the fixes that actually address it.
9 min read
AI for Developers
When to Reach for an AI Assistant — and When to Write It Yourself
AI is brilliant at some coding tasks and quietly costly at others. A practical map of where assistants save real time versus where they add hidden risk, so you delegate on purpose.
9 min read
AI for Developers
Why AI Invents APIs That Don't Exist, and How to Not Get Burned
AI assistants confidently reference functions, packages and flags that were never real. What causes these hallucinations, the security risk of "slopsquatting", and the verification habits that protect you.
9 min read
AI for Developers
Teaching a Kid to Code: What Works, What Backfires, and Why Scratch Isn't a Toy
A developer-parent's field guide: ages and stages, block languages versus 'real' code, the AI question, and how to keep the spark alive past the first hour.
9 min read
AI for Developers
Using AI to Review and Test Code Without Outsourcing Your Judgement
AI is surprisingly good at spotting issues and drafting tests — and surprisingly good at missing the important ones. How to use it as a tireless first-pass reviewer while staying the final authority.
9 min read
AI for Developers
Prompting an AI Assistant Like an Engineer, Not a Wizard
Vague prompts get vague code. The concrete habits that make AI coding assistants reliably useful — context, constraints, examples — and the trap of trusting fluent wrong answers.
9 min read
AI for Developers
Using AI Coding Assistants Without Shipping Their Mistakes
AI assistants write plausible code fast — including plausible bugs. A practical approach to getting the speed without lowering your standards or your security.
9 min read