AI Research Playbook — Full Notes & Workflow
A consolidated notes page that turns raw exploration into a repeatable, profitable AI research process.
0) Purpose & Outcomes
Goal: Turn AI-powered research into monetizable outputs (content, tools, services, products) while staying accurate and efficient.
Core outcomes you should ship:
- Actionable briefs, outlines, and scripts (YouTube, posts, newsletters)
- Evidence-backed comparison tables (CSV/Sheets) and choosers
- Research-driven lead magnets and mini-products
- Slides, mind maps, and summaries for fast decision-making
1) Principles & Mindset
- Stack, not a single tool. Use best-of-breed tools for different steps; stop asking one model to do everything.
- Specific beats generic. If a one-line prompt can generate it, the output is a commodity. Seek angles and evidence others don’t.
- Categorical reference. Learn by sorting: create “boxes” (categories) first, then place findings inside.
- Spin for your niche. Constantly reframe findings for the audience and monetization path you care about.
- Research → Decisions → Assets. Entrepreneurs get paid to decide and ship—not to read forever.
- Run lean, not cheap. Pay where it compounds into quality or time savings (e.g., multi-agent wide research).
2) The Research Stack (Roles & When to Use)
Conversational / Ideation
- ChatGPT (general) – fast brainstorming, framing, prompt drafting, outline first-pass.
Web-validated research
- Gemini Deep Research – systematic crawl on a focused question; transparent sources.
- Perplexity / Web Q&A – quick source-backed answers and passages.
Agentic, multi-branch exploration
- Manus AI (“Wide Research”) – multi-agent, multi-path exploration; builds large, source-rich reports/tables; slower, pricier, but high leverage.
Second brain / Synthesis
- NotebookLM – ingest notes, produce study guides, timelines, mind maps, audio dialogs; excellent for learning & content reuse.
Visualization / Packaging
- Spreadsheets/CSV, Notion/Docs, slide generators, prompt-to-slides.
Rule of thumb: Start with ChatGPT to frame the question → run Deep or Wide research where needed → synthesize in NotebookLM → package deliverables.
3) Four-Phase Workflow (Idea → Output)
Phase A — Define & Aim
- Audience & intent: Who is this for? What problem are they trying to solve right now?
- Artifact & monetization: What will you ship? (brief, chooser, guide, tool) How does it make money?
- Success criteria: What decision will this research enable?
Phase B — Discover & Collect
- Use ChatGPT to map the space (terms, players, criteria, unknowns).
- Run Gemini Deep Research for systematic, source-backed evidence.
- Use Manus AI (Wide Research) when you need breadth + depth + structured outputs (e.g., CSV of 25–150 items with attributes, quotes, and sources).
- Save sources, quotes, tables. Tag with intent and monetization relevance.
Phase C — Synthesize & Structure
- Build comparison matrices (columns: entity, method, metrics, obstacles, outcomes, links, quotes).
- Use NotebookLM to generate study guides, timelines, mind maps, and FAQs from your corpus.
- Extract insights (patterns, thresholds, “if X then Y” rules) and angles (hooks, spins, counterintuitive takes).
Phase D — Ship & Monetize
- Turn insights into: scripts, posts, carousels, cheatsheets, calculators, choosers, or sales letters.
- Add CTAs, affiliate links, or service offerings aligned to the research.
- Track results → iterate prompts, criteria, and packaging.
4) Tool Deep Dives (Practical Use)
ChatGPT (Conversational & Framing)
Use for: problem framing, question trees, prompt engineering, outline drafts, transforming notes into copy. Deliverables: briefs, outline v1, prompt libraries, draft scripts, sales letters. Pitfalls: flat/uncited facts, generic takes. Fix with better constraints and follow-up with source-backed tools.
Starter prompts:
Frame a research plan: audience, desired decision, monetization path, and success criteria. Ask me for any missing constraints.
Turn these raw notes into a 500-word executive brief with 5 key findings, 3 risks, and next actions.
Gemini Deep Research (Focused Evidence)
Use for: deep dives on a single question; transparent browsing trail; faster than wide research. Deliverables: curated source set, reasoned summaries, claim verification. Tips: specify evidence requirements; ask for contradictions; require quotes + links.
Starter prompt:
Run deep research on: “Best laptops for students who do light video editing under $900.”
Return a table with model, CPU/GPU, RAM, weight, battery claims, and 3 trusted sources per pick.
Highlight trade-offs and recurring failure points.
Manus AI — “Wide Research” (Multi-Agent)
Use for: broad+deep sweeps that output structured tables, quotes, and linked sources across many entities (25–150+). Deliverables: CSV/Sheets, longform reports, slide decks with citations. Notes: May take longer (minutes → hours). Worth it when you need breadth, structure, and quotable evidence. Good for competitive landscapes, method comparisons, or historical analyses.
Starter prompt:
Use Wide Research to identify 50+ ways marketers lower paid ad costs for podcasts.
For each: tactic, mechanism, example link, expected lift range, constraints, proof sources, and 1-sentence playbook.
Export CSV and a summary slide deck.
NotebookLM (Second Brain)
Use for: ingesting research outputs to generate mind maps, timelines, study guides, audio dialogues, FAQs. Deliverables: “learning pack” for you/team; assets you can repurpose into content. Tips: paste entire reports/CSVs; ask for mind maps by theme; generate a briefing doc before writing.
5) Working Concepts & Heuristics
- Commodity filter: If a first-try prompt yields what everyone else has, keep digging until you have a table, a rule, a framework, or a benchmark others don’t.
- Hook from insight: Lead with patterns, thresholds, and contradictions (e.g., “Among 150 cases, walking ≥1 mile/day appeared in 72% of ≥25 lb losses”).
- Spin for market: Reframe generic findings to your buyer (parents, gym owners, editors, etc.).
- Evidence cadence: claim → source → quote → implication.
- Stop rule: Ship once your matrix stops changing with new sources (diminishing returns).
6) Examples (from the workflow)
A) Celebrity Weight Loss → Pattern Report
- Move from listicles to a matrix: person, timeline, method (diet/exercise), challenges, quotes, sources.
- Derive actionable patterns (e.g., activity minimums, protein targets, adherence tactics).
- Monetize: checklist/guide, mini-course, partner offers (coaching, trackers), content series.
B) Student Laptops for Light Editing → Chooser Tool
- Criteria: price ceiling, CPU/GPU floors, RAM, weight, battery, ports, warranty.
- Build a chooser (inputs → ranked picks) + affiliate links.
- Monetize: evergreen posts, comparison sheets, email capture via buyer’s guide.
C) Philosophers → Modern Angles
- Crossovers: who influenced whom; where ideas clash.
- Apply to AI: Socratic method prompts, logic/rationalism for prompt debugging, “Plato’s Cave vs. the Metaverse.”
- Monetize: thought-leadership content, course, newsletter series.
D) Lowering Ad Costs (Podcast Promo)
- Wide research: 25–50 tactics with constraints, examples, and expected lift.
- Build a playbook + calculator to prioritize by budget and channel.
- Monetize: lead magnet → audit/service; productized consulting.
E) Niche How-To (e.g., Raising Quail)
- Feasibility check: enough sources? If yes, table the setup/care/ROI/law/compliance.
- Package as step-by-step with sourcing and seasonal/locale notes.
7) From Research → Revenue (Packaging Menu)
- Content: briefs, posts, carousels, long videos, shorts.
- Data: CSV databases, choosers, calculators, benchmarks.
- Products: paid guides, courses, templates, swipe files.
- Services: research-as-a-service, audits, strategy sprints, due diligence.
- Funnels: intent-tailored lead magnets → email sequences → offers.
8) Prompt Library (Plug-and-Play)
A. Frame & Scope
Act as a research lead. Define audience, decision to enable, monetization path, and a minimal evidence plan (top sources, data to collect, what *not* to do). Ask 5 scoping questions.
B. Wide Research (Manus AI)
Use Wide Research to produce a 100-row table on <topic>.
Columns: Entity/Method, Mechanism, When it Works, Constraints, Metrics, Example Link, 2–3 Quotable Lines with sources. Export CSV + summary slides.
C. Deep Research (Gemini)
Deep research: “<question>”. Require 6+ diverse high-quality sources.
Return an executive brief + a table with comparable attributes and direct quotes with citations.
D. Synthesis → Insight Matrix
From these notes, extract patterns, thresholds, contradictions, and 5 ‘if/then’ rules. Produce a matrix and 10 hooks for content.
E. Packaging
Turn the insight matrix into: (1) a 7-slide deck, (2) a 400-word newsletter, and (3) a checklist lead magnet with CTA.
F. Sales Letter from Research
Using the pains, constraints, and proof collected, write a direct-response sales letter for <offer>. Include specific problem statements, mechanisms, proof points, FAQs, and a risk-reversal.
G. Socratic Method Prompt
Act as a Socratic coach. Ask me 12 clarifying questions about <goal>. Challenge assumptions, request evidence, and help me derive a testable plan.
9) Templates
Insight Table (CSV):
Item | Category | Method/Mechanism | Inputs | Constraints | Outcomes/Thresholds | Failures | Sources | Quotes |
---|
Claim Verification:
- The claim (1 sentence)
- Source list (diverse, reputable)
- Direct quotes (max 25 words each)
- Counterevidence
- Verdict (+ caveats)
Deliverable Checklist:
10) Accuracy, Ethics & Pitfalls
Best practices
- Cite and quote. Distinguish fact vs. inference.
- Use multiple tools to triangulate.
- Track dates; watch for stale info.
- Keep raw notes separate from polished outputs.
- Attribute ideas; avoid plagiarism.
Common pitfalls
- Flat, generic outputs; no edge or evidence.
- Overfitting to one tool; skipping verification.
- Blindly trusting numbers or affiliate lists.
- Researching forever without shipping.
11) 7‑Day Research Sprint (example)
Day 1: Define audience, decision, success criteria. Draft question tree.
Day 2: Deep research pass; collect 10–15 top sources.
Day 3: Wide research pass; export CSV.
Day 4: Synthesize into insight matrix and hooks.
Day 5: Package: brief + deck + lead magnet.
Day 6: Publish; email + social + video.
Day 7: Measure; refine prompts/criteria; plan next sprint.
12) Quick Reference
- Use ChatGPT to frame problems and create prompts.
- Use Gemini (Deep) for focused, evidence-backed answers.
- Use Manus (Wide) for structured breadth with sources.
- Use NotebookLM to learn, map, and repurpose.
- Always ship a brief + table + one monetized asset.
Master the stack, respect evidence, and ship assets that make decisions—and money.