Prompt Engineering

1M-token context windows exist, but quality degrades long before that limit. Context engineering is the discipline of using context windows effectively — what to include, what to summarize, what to retrieve fresh, and the patterns that keep quality high as context grows.

Production prompts are not 'tell the AI what you want'. They are a layered system — stable instructions, dynamic context, per-call variables — managed like code. The architecture, the patterns, and the discipline that distinguishes production from prototype.

Reasoning models are not fast models with extra steps. They reward different prompting, ignore some conventional patterns, and have their own pitfalls. A practical guide to working with them well.

Three reasoning techniques that genuinely improve AI output on hard problems — and the cost-benefit math of using them. With concrete prompts, side-by-side comparisons, and the gotchas modern reasoning models introduce.

Once you use AI seriously, you write the same kinds of prompts over and over. A practical system for building, organising, and sharing a prompt library — what to capture, how to version, and what infrastructure to use.

Ten reusable patterns — not one-off prompts, but reusable shapes — that cover the vast majority of practical AI use at work. Memorise these and you have an entire prompt-engineering toolkit.

Spend ten minutes once and stop re-explaining yourself to ChatGPT every conversation. A practical guide to custom instructions, memory, and projects across the major AI tools.

When you cannot describe what you want, show it. Few-shot prompting — giving the model 2-3 examples of the output style — is the single fastest way to get consistent, on-brand results.

The single biggest gap between mediocre and good AI output is what happens after the first response. A workflow for turning any first draft into a real one through iteration — not by re-prompting from scratch.

A five-part template that lifts answer quality from \"meh\" to \"useful\" on any prompt, in any AI tool. With worked examples, common mistakes, and the order that matters.

Five real prompts shown in their weak version and their better version, with a short note on what changed. The fastest way to upgrade your AI output without learning any jargon.

LangChain. Lance Martin's framework — write, select, compress, isolate — with concrete examples of when to summarise an action history, when to offload state to files, and when to spin up sub-agents purely to protect the parent's context. Maps almost directly onto the article's section on managing 1M-token windows in practice.

Chroma. Kelly Hong walking through Chroma's research on 18 models — why needle-in-haystack scores are misleading, how performance degrades with ambiguity and distractors, why even simple string-repetition tasks degrade past 500 tokens. Short, evidence-based, and exactly the case the article needs you to take seriously before getting to the engineering moves.

Anthropic. A live build session by Anthropic's Applied AI team on an insurance-claims prompt — they start with a vague instruction and iterate to something a developer would actually ship, showing the kind of revisions the article describes for the system and developer layers. Watch this before re-reading the article's checklist on examples, output structure, and refusal handling.

Anthropic. Four Anthropic prompt engineers (research, alignment, applied, developer relations) talking at length about what they actually do day to day — how they edit prompts under pressure, how they think about "honesty" in instructions, when XML scaffolds help, when they don't. The article's layered model maps cleanly onto how they describe the work; this is the best way to hear that mental model out loud.

OpenAI. Three minutes with the o1 team describing the moment the model started questioning its own reasoning during RL training. Useful as the primary source for the article's claim that the chain-of-thought is now happening inside the model, not in your prompt.

AI Explained. The clearest non-hype explanation of why o1 (and by extension o3, R1, Claude extended thinking) is trained for correctness instead of next-token prediction, and what that means for how you talk to it. Watch this and the article's "stop saying think step by step, start writing the spec" advice stops feeling arbitrary.

Yannic Kilcher. Yannic walks through the original Tree-of-Thoughts paper from Princeton/DeepMind — the generate-score-prune-backtrack loop, why it lifts Game of 24 accuracy from 4% to 74%, and where the technique stops being worth its 10× token cost. Useful precisely because it forces you to see ToT as a search algorithm, not a magic prompt prefix.

Anthropic. Two members of Anthropic's Applied AI team build up a real-world prompt live, layering in role, structured sections and explicit thinking steps. It's the most credible "how to use chain-of-thought without ritualizing it" demo on YouTube, and it makes the article's "stop pasting Let's think step by step into everything" point concrete.

Lawton Learns. A concrete, copy-pasteable C.R.A.F.T. meta-prompt (Context, Role, Action, Format, Target audience) that the article references in spirit. Useful as the first entry in a new library — a prompt whose job is to generate well-structured prompts for everyone else.

Tina Huang. Tina compresses Google's full prompting course into the two things a library actually stores: a five-part task/context/references/evaluate/iterate framework, and four named iteration moves to tighten a prompt over time. She also walks through agent and chain-of-thought patterns, so the same video doubles as the entry point for the more advanced templates you'll add to the library later.

Tina Huang. The prompting chapter (02:30–09:20) lays out two stackable mnemonics — "tiny crabs ride enormous iguanas" (task, context, references, evaluate, iterate) and "ramen saves tragic idiots" (revisit, separate, try analogous, introduce constraints) — that map cleanly onto the iterate-and-refine patterns in the article. Useful when a single-shot formula isn't getting you there.

Jeff Su. Six-part formula — task, context, exemplars, persona, format, tone — explained with the same office-work examples the article uses (workout plans, resumes, internal emails). Watch this first; almost every pattern in the article is a focused application of one of these six components.

Andrej Karpathy. Skip to the "ChatGPT memory, custom instructions" chapter at 1:53:29. Karpathy shows what he actually has in his own custom instructions, talks about when he tells ChatGPT to remember a fact versus letting it auto-capture, and notes the trade-offs — closer to the article's "set it up once with intent" framing than most beginner walkthroughs.

Tina Huang. The personalization, memory, and custom instructions panels are walked through inside a wider tour of every ChatGPT feature, which is the right framing — these settings only pay off when you also know what they're configuring (projects, GPTs, scheduled tasks, voice, etc.). Use it as a guided tour while you actually open Settings → Personalization in another tab.

Anthropic. Examples are one of the ten ingredients Anthropic walks through here, and you can watch them get added to a real customer prompt and change the model's output. Useful if you want to see few-shot used inside a fuller prompt rather than as an isolated trick.

freeCodeCamp.org. Skip to the "Zero shot and few shot prompts" chapter at 31:20 for a clean, demoed comparison of the same task with no examples versus a few — exactly the contrast the article builds the chapter around. The earlier sections on prompt mindset and best practices are useful background if you've come straight from the article without the wider context.

Andrej Karpathy. Karpathy walks through his actual day-to-day use across ChatGPT, Claude, Gemini, Grok, and Perplexity, screen-sharing live conversations rather than polished one-shots. You see when he reaches for a follow-up, when he switches models mid-thread, and when he abandons a line and starts over — exactly the iterative loop the article describes, modeled by someone who builds these systems for a living.

Tina Huang. A condensed run through Google's prompt engineering course covering frameworks for clarity, iteration methods, and worked examples for emails, summaries, and analysis. Useful if you want a second pass that maps the same structural ideas onto the kinds of tasks you actually hit during a workday.

Anthropic. Two members of Anthropic's Applied AI team build up a single prompt for a real customer use case (analyzing Swedish car accident reports), starting from a one-liner that fails and adding role, task context, dynamic content, examples, and final emphasis one layer at a time. It is the cleanest demonstration on YouTube of the exact "role + context + task + constraints + format" stack the article describes, with the failure modes shown along the way.

freeCodeCamp.org. A longer, more patient course that earns its length. The chapters on best practices, zero-shot vs few-shot, and what hallucinations look like in prompted output are exactly the deepening you want once the article's five examples have clicked.

Tina Huang. Tina condenses Google's official Prompting Essentials course into the five-part TCREI structure (Task, Context, References, Evaluate, Iterate) with worked examples on each. After the article shows you five rewrites, this hands you the recipe Google itself teaches employees, so you can do the rewrite on your own prompts without copying ours.