The Idea
When AI writes a long response, it generates one word at a time from start to finish. But many responses have sections that don't depend on each other — the paragraph about physical health doesn't need to wait for the paragraph about mental health to be written first.
Skeleton of Thought exploits this. First, ask AI to create a quick outline (the "skeleton"). Then expand each outline point into a full section simultaneously. Because the sections are independent, they can all be written at the same time — cutting wait time roughly in half.
Building Blocks
This composition combines:
One Thing at a Time Structure the OutputIt uses structured output to generate an outline, then applies "one thing at a time" to each section independently — with the twist that all sections expand in parallel.
See It in Action
Question: "What are the benefits of regular exercise?"
2. Mental health benefits
3. Social and lifestyle advantages
4. Long-term disease prevention
Why This Works
Standard AI generation is like writing a book from first word to last — every word waits for the one before it. But most long responses have natural sections that are independent of each other. Writing them in parallel is like having four authors each write one chapter simultaneously.
The outline stage is key: it ensures all sections are planned coherently before any are expanded. This prevents overlap and gaps. And because each expansion focuses on just one point, the individual sections tend to be more focused and detailed than they would be in a single long generation.
The Composition
Ask AI for a quick outline first. Then expand every point into a full section simultaneously. Assemble the parts. You get structured, detailed content in roughly half the time.
How to Apply This
- Ask: "Give me a 4-5 point outline for [topic]. Just the main points, a few words each."
- Then for each point, start a separate conversation: "Expand point [N] into a detailed paragraph about [topic]."
- Run all the expansion requests at the same time (use multiple browser tabs, or AI tools that support parallel requests)
- Combine the expanded sections into your final document
When to Use This
- • Long-form content: articles, essays, reports, explanations
- • Multi-aspect analysis where each aspect is independent
- • When speed matters and you need results quickly
- • Any response that naturally has a list or section structure
- • When you want better organization — the outline forces clear structure
When to Skip This
- • Math problems — each step depends on the previous one; can't be parallelized
- • Narrative writing — stories need flow and coherence between sections
- • Short responses — the outline overhead isn't worth it for a quick answer
- • Tightly connected reasoning — if section 3 must reference section 2, they can't be written independently
How It Relates
This is the parallel-execution cousin of Chain It. Where Chain It runs steps sequentially (output of step 1 feeds into step 2), Skeleton of Thought runs the expansion steps simultaneously because they're independent. It also complements Structure the Output — that technique asks for organized formatting, while Skeleton of Thought uses structure as a strategy for faster, better generation.