Why AI gets things wrong, organized by root cause — a diagnostic field guide for engineers, PMs, and serious users who want a causal map instead of an anecdote.
Most writing about AI defects circulates anecdotally — one-off complaints about hallucinations, sycophancy, or confidently wrong answers. Readers experience these failures constantly but lack a systematic frame for understanding why they happen. Without that frame, every defect feels like a surprise, a one-off bug, or a personal mistake. This collection organises defects by root cause rather than by symptom — data, architecture, training, reasoning, performed judgement, conversational drift, and the user's own contribution as amplifier — and ends with the judgement layer for working with the tool well. The pair to Understand Harness Engineering: that one explained how the agent works, this one explains how it fails. Honest, clinical, calm. Not contrarian, not doomer — calibrated for the working user who wants the vocabulary to use AI competently because they understand its limits.
Certificates, credits, and a record of your progress.
Reach 60% of its base credits to earn a Certificate of Completion.
View example certificateCredits accumulate in your learning tracks, building a subject-specific record.
See your credits, reading progress, quiz performance, and available exam results.
View a sample reportOne session. One clear takeaway.Each short session develops one distinct idea with a clear takeaway. Together, the sessions show how those ideas connect into a structured understanding of the topic.
Defects are not bugs but structural consequences — and the four sources (data, architecture, training, interaction) that organise the rest of the collection.
Defects are not bugs but structural consequences — and the four sources (data, architecture, training, interaction) that organise the rest of the collection.
What the training corpus bakes in, and what token-by-token generation cannot do no matter how the prompt is shaped.
What the training corpus bakes in, and what token-by-token generation cannot do no matter how the prompt is shaped.
The side effects of RLHF and reward modelling, and the failure modes that pass for reasoning but aren't.
The side effects of RLHF and reward modelling, and the failure modes that pass for reasoning but aren't.
Performed judgement that looks like real judgement, and the drift modes that emerge over a long thread.
Performed judgement that looks like real judgement, and the drift modes that emerge over a long thread.
The prompt habits and reading habits that turn the model's defects into the user's defects.
The prompt habits and reading habits that turn the model's defects into the user's defects.
The judgement layer — how a defect-aware user prompts, verifies, and routes work differently.
The judgement layer — how a defect-aware user prompts, verifies, and routes work differently.
“A diagnostic map of AI's defects, organised by root cause rather than by symptom.”
What sets this collection apart is its refusal to treat AI's failures as anecdote. It opens the model, names where each defect comes from, and does the harder editorial work of separating defects of correctness from defects of performed judgement. Honest about the user's contribution as well as the model's. The pair to the harness-engineering collection — that one explained how the agent works, this one explains how it fails.
Working with AI is a LearningFirst editorial line publishing field-level skills for working professionals adapting to AI in their daily craft. The aim is durable practice, not feature coverage — the patterns that transfer across tools and outlast any one model release.
Ranked by semantic similarity
AI didn't just change what students can do. It changed what teaching is for.
AI can think for you. The question is whether you still want to.
AI keeps getting stronger — here is what to build that it structurally cannot replicate, and how to compound it
