Chunk

CHUNK

CHUNK is the prompt trick you use when your source material is a mess, a monster, or both. Instead of asking ChatGPT to deal with one giant blob of text, you tell it how to break the material apart, label the parts, check that it understands them, and keep the original sequence intact. Basically, it turns a wall of text into something your brain can survive.

If you have notes, transcripts, meeting dumps, research pasted from six tabs, or a document that looks like it lost a fight with formatting, CHUNK helps.

What the template means

Letter or word	Prompt ingredient	What it does
C, Cut	Break the material into smaller pieces	Prevents overload and makes the content manageable
H, Headline	Create a short heading for each piece	Gives each chunk a quick summary at a glance
U, Understand	Check the meaning of each piece	Makes the model interpret the section instead of just slicing it
N, Name	Assign a useful label or category	Helps organize chunks for later use
K, Keep order	Preserve the original sequence	Keeps the structure and flow of the source material

Why this template works

A lot of weak prompts fail for one simple reason: they ask the model to process too much material at once, with no structure.

CHUNK fixes that by giving ChatGPT a job sequence:

Split the material.
Summarize each part.
Show understanding.
Label the part.
Preserve the order.

That matters because large documents are rarely just “information.” They usually have progression. A meeting goes from intro, to problem, to debate, to decision. A training transcript goes from basics, to examples, to warnings. If the order gets scrambled, the meaning often gets scrambled too.

CHUNK is especially useful in the “working with material” stage of prompting. This is where you’re no longer asking random one-off questions. You’re asking ChatGPT to process actual content, like notes, transcripts, reports, or pasted documents. At this stage, structure is not optional. It is the difference between “helpful assistant” and “confident intern who sorted everything wrong.”

3 sample prompts

Sample prompt 1: Formal version

CUT: Divide the transcript below into 6 logical chunks based on topic changes.
HEADLINE: Give each chunk a short headline of 3 to 6 words.
UNDERSTAND: For each chunk, write 1 sentence explaining the main idea in plain English.
NAME: Assign each chunk a category label such as Introduction, Problem, Evidence, Recommendation, or Next Steps.
KEEP ORDER: Present the chunks in the same order they appear in the original transcript.

Material: [Paste transcript here]

How the template shows up:

C, Cut: The prompt explicitly asks for 6 logical chunks.
H, Headline: Each chunk must get a short headline.
U, Understand: The model must explain the main idea of each chunk in plain English.
N, Name: Each chunk gets a category label.
K, Keep order: The prompt clearly says to preserve the original sequence.

Why it improves the prompt: This version is excellent when you want reliability and clarity. Nothing is implied. Every step is stated. That reduces the chances of ChatGPT giving you random sections, vague summaries, or a reorganized output you never asked for. It is formal, yes, but formal prompts are often the least dramatic and most useful. Like a good spreadsheet.

Sample prompt 2: Semi-structured version

I pasted rough project notes below.
Please break them into clear sections by topic, give each section a short headline, explain what each section is mainly about, and label each one with a practical name like risk, timeline, budget, or decision. Keep everything in the same order as the original notes.

Notes: [Paste notes here]

How the template shows up:

C, Cut: “Break them into clear sections by topic” tells the model to chunk the material.
H, Headline: “Give each section a short headline” covers the headline step.
U, Understand: “Explain what each section is mainly about” asks for interpretation, not just splitting.
N, Name: “Label each one with a practical name like risk, timeline, budget, or decision” handles naming.
K, Keep order: “Keep everything in the same order” preserves the source flow.

Why it improves the prompt: This version feels more natural while still being structured. It works well for everyday tasks like meeting notes, planning docs, or messy brainstorming sessions. You are not forcing a rigid format, but you are still giving enough guidance to avoid a vague summary blob. That is the sweet spot for many real-world prompts.

Sample prompt 3: Natural version

Here are my class notes from a workshop. Split them into smaller sections, give each one a quick title, tell me what each part is saying in simple terms, call each section something useful, and keep the same order so I can study them without getting lost.

How the template shows up:

C, Cut: “Split them into smaller sections” is the chunking instruction.
H, Headline: “Give each one a quick title” creates headlines.
U, Understand: “Tell me what each part is saying in simple terms” checks understanding.
N, Name: “Call each section something useful” asks for naming or categorization.
K, Keep order: “Keep the same order” protects the original structure.

Why it improves the prompt: This version proves you do not need to sound like a compliance officer to write a strong prompt. It is casual, but still specific. The model gets clear instructions on what to do with the material, how to present it, and what not to mess up. For study notes, transcripts, or long articles, this makes the output much easier to review later.

Warnings and extra tips

CHUNK is great, but it is not magic. A few things to watch for:

Do not skip the source material. CHUNK works on actual content. If you forget to paste the notes, transcript, or document, the model will cheerfully improvise. That is not what you want.
Be realistic about chunk size. If you say “break this into chunks” without saying how many or by what logic, the model may create sections that are too broad or too tiny.
“Understand” matters. Without the U step, you might get clean-looking chunks that are still shallow. A nice heading is not proof of comprehension.
Naming can be too vague. If you want labels like “Problem,” “Action Item,” or “Decision,” say so. Otherwise you may get labels that sound polished but are not useful.
Keep order only when order matters. For transcripts, lessons, and timelines, order matters a lot. For some brainstorming sessions, you may actually want the model to reorganize by theme instead.
This template is not enough for every task. If you also need a specific output format, audience, tone, or final transformation, add those instructions too. CHUNK helps process material, but it does not replace every other prompt ingredient.
Messy input still affects the result. If your source notes are full of missing context, abbreviations, or half-finished thoughts, the model may guess. You may need to clean the material a little first, or ask it to flag unclear sections.

A handy tip: ask for a table output if you want to review chunks quickly. For example, you can request columns like “Chunk Number,” “Headline,” “Main Idea,” and “Category.” Suddenly your chaos looks suspiciously organized.

Quick summary

CHUNK helps you break big material into smaller, labeled, understandable pieces without losing the original order:

Cut the material
Headline each part
Understand what each part means
Name it clearly
Keep order

When a document is too long, too messy, or too confusing, CHUNK gives ChatGPT a process instead of a panic attack.

Try it next with something real: a meeting transcript, workshop notes, a long article, or your own chaotic brain-dump document. Paste it in, use CHUNK, and see how much easier the material becomes to work with.