🛰️ Research and discovery
The commands in family 02: structured investigation, broad scanning, discovery of the non-obvious, and accurate reading of trends and weak signals.
🔬 /research — investigate systematically
The protocol for /research: defines questions, collects evidence, synthesizes findings, and wraps up with gaps and next steps. The difference from "searching the internet" is the wrap-up contract—the answer only ends when it states what’s still missing.
📋 The four required deliverables
- Explicit research questions (not a vague topic)
- Evidence collected with its source identified
- Synthesis of the findings—not a collage of excerpts
- Gaps + concrete next steps
💡 Practical tip
If the answer doesn't end with gaps, run it again with "/research … declare gaps". Without enforcing the contract, the command tends to pretend it's complete.
🎯 Copy and run
/research sobre; perguntas-chave primeiro, depois evidências, síntese e lacunas
How to verify: four blocks in contract order—and no key question answered evasively.
🕳️ /deepresearch — dig deep through triangulation
Expands sources, triangulates evidence, compares interpretations, and documents relevant conflicts between them.
New here? Triangulation is confirming the same finding through independent paths (primary source + your own data + an expert). If three sources cite the same original report, they aren’t three pieces of evidence — they’re one.
Depth × breadth: each step costs time — go deeper only where the value justifies it.
⚠️ Actual cost
Deepresearch is the most expensive command in the family. Use it after /scan has identified WHERE it’s worth digging in—never as your first response to any question.
🎯 Copy and run
/deepresearch sobre; triangule fontes e documente conflitos entre interpretações
How to verify: each central finding is supported by at least two independent sources; disagreements remain visible rather than being smoothed over.
📡 /scan — scan broadly without getting bogged down
Searches for emerging signals, players, solutions, developments, and themes without going too deep too soon. It’s the quick map before the expedition.
List actors and movements
Who exists, who is moving, what has changed recently—without judging yet.
Mark candidates for further review
The natural output of the scan is a prioritized shortlist for /research or /deepresearch.
Resist premature deep dives
If you're already reading 80-page PDFs in the scan, you've gone beyond the command—go back.
🎯 Copy and run
/scan sobre; liste atores, movimentos e os 3 melhores alvos para aprofundar
How to verify: broad list + justified shortlist. No deep analysis disguised as a scan.
💎 /discover — find what isn’t obvious
Search patterns, adjacencies, combinations, and underexplored areas. While /scan looks at what's already moving, /discover looks for what no one has connected yet.
✓ Discovery questions
- ✓"Could someone who solves a neighboring problem solve mine?"
- ✓"What combination of two existing features has no one tried?"
- ✓"Which adjacent audience has the same pain point and more money?"
✗ False discoveries
- ✗A combination the whole industry has already tested and abandoned—for a documented reason.
- ✗“Empty” space that’s empty because there’s no demand and no path to it.
🎯 Copy and run
/discover oportunidades não óbvias em; proponha combinações improváveis com evidência de viabilidade
How to verify: each opportunity comes with the reason it remains unexplored — and the cheap test that would disprove the hypothesis.
🧩 /gaps — discover what’s missing
Identifies missing questions, required data, risks, exceptions, and overlooked opportunities. It’s the family’s skeptical command: it assumes every piece of material has gaps—and goes looking for them.
🔍 The five /gaps hunts
- •Missing question — what no one asked that would change the conclusion.
- •Required data — the information without which everything is an educated guess.
- •Risk not mentioned — what the material would rather not see.
- •Silenced exception — cases where the rule failed and disappeared from the text.
- •Overlooked opportunity — the obvious alternative use no one proposed.
🎯 Copy and run
/gaps em; priorize as 3 lacunas que mais ameaçam a conclusão
How to verify: gaps ranked by impact on the conclusion — not by how easy they are to resolve.
📈 /trends + /signals — hype vs. change
The horizon pair: /trends classifies the current movement into four types; /signals looks for early signs that foreshadow larger changes.
📊 The /trends taxonomy
Hype without economic backing. It dies on its own.
A real curve with unit economics that work.
Small today, with a mechanism to grow.
Permanently changed the rules of the game.
💡 Practical tip
Before betting on a trend, require the command to answer: "what would make this just a fad?" If the answer is convincing, you just saved a year.
🎯 Copy and run
/trends /signals sobre; classe em modismo/crescimento/sinal/estrutural com evidência
How to verify: each trend classified WITH the criterion that put it there — not by market intuition.
🔮 /future — build plausible scenarios
Projects alternative scenarios to test whether your plans hold up outside an optimistic world. A scenario is not a prediction: it is a robustness lab.
Choose the key uncertainties
Two variables that actually change the game—not ten decorative ones.
Describe 3–4 coherent scenarios
Each one has observable triggers: “if X happens, we’re in this scenario.”
Test the plan on all of them
A plan that fails in two out of four scenarios needs a hedge — not faith.
🎯 Copy and run
/future cenários paraem 24 meses; marque gatilhos observáveis de cada cenário
How to verify: distinct scenarios (not variations of the same one), each with a clear leading indicator.
🗺️ /benchmark + /landscape — references and ecosystem
Close the module with the surrounding context: /benchmark compares your performance against real benchmarks; /landscape maps the entire ecosystem — players, substitutes, suppliers, and forces.
✓ Honest benchmark
- ✓Comparable metrics (same definition of "active" or "churn").
- ✓Reference chosen for relevance, not flattery.
✗ Vanity benchmark
- ✗Compare against the worst competitor available to make it look good.
- ✗Mixing metrics with different definitions in the same table.
🎯 Copy and run
/benchmark /landscape para; inclua substitutos indiretos e diga onde estamos posicionados
How to verify: map with direct players AND substitutes, and your stated position with criteria — not in the center for convenience.
Quick self-check (optional): before spending hours on /deepresearch, which command should come first?
🎓 Module Summary
Next Module:
2.2 - Analysis and synthesis — turn this research's findings into models, causes, and simplified decisions.