https://plainlogic.org/blog/ Technology news and explainers in plain words, from the PlainLogic lab. en https://plainlogic.org/blog/what-are-ai-evals/ https://plainlogic.org/blog/what-are-ai-evals/ AI evals test models the way unit tests test code: inputs in, graded outputs, a score. How they work and why they matter. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-are-transformers/ https://plainlogic.org/blog/what-are-transformers/ Transformers power modern AI. What the architecture is, how self-attention works, and why it changed everything — in plain language. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-a-context-window/ https://plainlogic.org/blog/what-is-a-context-window/ A context window is an AI's working memory — the text it can consider at once. When chats outgrow it, earlier parts drop out. Here's why. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-a-system-prompt/ https://plainlogic.org/blog/what-is-a-system-prompt/ A system prompt is the hidden instruction set shaping every AI reply. What it is, who writes it, and why AIs answer differently. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-an-agentic-workflow/ https://plainlogic.org/blog/what-is-an-agentic-workflow/ An agentic workflow lets an AI plan, act with tools, and observe results in a loop. Learn how the pattern behind AI agents works — and when not to use it. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-chain-of-thought/ https://plainlogic.org/blog/what-is-chain-of-thought/ Chain-of-thought means asking AI to show its work before answering. 'Think step by step' measurably improves reasoning — here's why it works. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-distillation/ https://plainlogic.org/blog/what-is-distillation/ Model distillation trains a small, fast AI model to imitate a big one's judgment. Learn how the teacher-student trick behind on-device AI actually works. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-fine-tuning/ https://plainlogic.org/blog/what-is-fine-tuning/ Fine-tuning is extra training that teaches an AI your style, format, or behavior — not new facts. How it differs from prompting, explained. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-grounding/ https://plainlogic.org/blog/what-is-grounding/ AI grounding connects answers to real sources you can check. How grounding reduces hallucinations, in plain language. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-inference/ https://plainlogic.org/blog/what-is-inference/ Inference means running a trained AI model on new data. How it differs from training, and why it drives AI costs. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-multimodal-ai/ https://plainlogic.org/blog/what-is-multimodal-ai/ Multimodal AI processes text, images, audio, and video together. How it works, what it's good for, and its limits — in plain language. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-prompt-injection/ https://plainlogic.org/blog/what-is-prompt-injection/ Prompt injection smuggles hidden instructions into AI apps through data they read. How it works, why there's no real fix, and what defenses help. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-rlhf/ https://plainlogic.org/blog/what-is-rlhf/ RLHF trains chatbots on human preferences: people rank answers, a reward model learns taste, the model optimizes for it. The technique behind polite AI. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-supervised-learning/ https://plainlogic.org/blog/what-is-supervised-learning/ Supervised learning trains AI on labeled examples — inputs paired with right answers. How it works, where it's used, and why memorization isn't learning. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-temperature/ https://plainlogic.org/blog/what-is-temperature/ Temperature is the AI setting that controls randomness: low means predictable answers, high means surprising ones. Here's how it works. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-tool-use/ https://plainlogic.org/blog/what-is-tool-use/ Tool use lets AI models call APIs and software instead of only writing text. Here's how the tool-call loop works. AI Tue, 06 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-an-llm/ https://plainlogic.org/blog/what-is-an-llm/ An LLM is not a search engine or a person in a box. It is a machine that learned the shape of language and writes text one token at a time. Plain words. AI Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-are-tokens/ https://plainlogic.org/blog/what-are-tokens/ AI models don't read words the way you do. They read tokens: fragments of text sliced by a fixed rule. What tokenization means, in plain words. AI Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-is-rag/ https://plainlogic.org/blog/what-is-rag/ Retrieval-augmented generation gives an AI model documents to read before it answers. How RAG works, what it fixes, and where it still goes wrong. AI Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/what-are-embeddings/ https://plainlogic.org/blog/what-are-embeddings/ Embeddings turn words, images, and ideas into lists of numbers so a computer can compare meaning, not just spelling. The simple explanation of semantic search. AI Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/ai-agent-vs-chatbot/ https://plainlogic.org/blog/ai-agent-vs-chatbot/ A chatbot talks; an agent works. The real distinction between a chat interface and an agent loop — and why the two keep getting mixed up. AI Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/when-not-to-use-ai/ https://plainlogic.org/blog/when-not-to-use-ai/ An LLM adds uncertainty, cost, and failure modes to everything it touches. The plain-logic test for when ordinary code beats AI — and when AI earns its keep. HOW TECHNOLOGY WORKS Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/why-ai-hallucinates/ https://plainlogic.org/blog/why-ai-hallucinates/ Language models predict text, not truth. Why confident wrong answers happen, what causes them, and how to check the answers that matter. AI Fri, 02 Oct 2026 00:00:00 +0000 https://plainlogic.org/blog/prompt-anatomy/ https://plainlogic.org/blog/prompt-anatomy/ A useful prompt states four things: the task, the context, the constraints, and the output format. What works, what doesn't, and what no prompt can fix. AI Fri, 02 Oct 2026 00:00:00 +0000