Analysis Toolkit
Eighteen Questions, a thirteen-cell Project Canvas, and a unified risk model. A fixed set of questions for any new project.
After 72 chapters, what matters most is not how much knowledge you remember, but whether a set of questions comes to mind automatically when you encounter a new project. This page brings those questions together.
Twelve questions
Start with these twelve questions when looking at a Crypto project. Every chapter's “Takeaway Questions” come from here.
- 1What problem does it solve?
- 2Why do we need blockchains?
- 3Why do we need a token?
- 4Who is the user?
- 5Who is paying?
- 6Where does revenue come from?
- 7How does a token capture value?
- 8Who supplies liquidity?
- 9Who bears the risk?
- 10What is the incentive?
- 11Can the product still run if its token price goes to zero?
- 12Will users remain when subsidies stop?
| Question | What it asks | Chapters to practice with |
|---|---|---|
| 1 What problem does it solve? | Is a real person experiencing a real difficulty? | Chapters 1, 13, 25 |
| 2 Why does it need a blockchain? | What changes if you use a database instead? | Chapters 2, 3, 9 |
| 3 Why does it need a token? | Does the proposition still hold without the token? | Chapters 10, 20, 30 |
| 4 Who are the users? | Real users, or activity farmers and airdrop hunters? | Chapters 23, 28 |
| 5 Who pays? | Whose pocket does the money come from? | Chapters 17, 26 |
| 6 Where does revenue come from? | Fees, interest spreads, or subsidies? | Chapters 17, 26 |
| 7 How does the token capture value? | How do protocol earnings benefit holders? | Chapters 19, 20 |
| 8 Who supplies liquidity? | Is someone committing real money on the other side? | Chapters 14, 15 |
| 9 Who bears the risk? | Who loses money first when something goes wrong? | Chapters 16, 18, 24 |
| 10 What are the incentives? | What do the rules encourage people to do? | Chapters 20, 28 |
| 11 Can it run if the token goes to zero? | Distinguish a product from a speculative asset. | Chapters 26, 30 |
| 12 Will users remain when subsidies stop? | Distinguish real demand from purchased demand. | Chapters 17, 28 |
Questions 11 and 12 are the toughest. Most projects fail here: when token prices fall, incentives cannot keep flowing, users immediately vanish, and the failure is blamed on “bad market conditions.”
Six AI questions
When AI is involved, add six more questions:
- 13What cost does AI reduce here?
- 14Does AI really need a blockchain?
- 15Why does an agent need a wallet?
- 16What permissions does an agent have?
- 17Who bears the loss when AI gets it wrong?
- 18Which decisions must keep a human in the loop?
Questions 17 and 18 are the easiest to skip and the costliest to ignore. A system that executes automatically must establish who is responsible before something goes wrong.
The thirteen-cell Project Canvas
A fixed format for understanding a project in 30 minutes. Empty cells show where your risk lies. Do not fill them with guesses; mark them “Not Found.”
Whose problem is it solving, and what exactly is the problem?
Who is actually using it today?
What are the protocol-layer rules? Who can change them?
What does the user actually see and do?
What sector is it in, how large is the market, and where are we in the cycle?
Is there a token? What are its supply, unlocks, and uses?
Where does protocol revenue come from, and who receives it?
Who is doing it, what have they done, and can it be verified?
Who invested, at what price, and when does it unlock?
Is the community made of users or airdrop hunters?
Who else is doing the same thing, and why will this one win?
What will change its position in the next three months?
What single event is most likely to reduce its value to zero?
See chapter 25 for instructions. A complete canvas takes about 30 minutes. If you spend more than an hour and cannot fill even half of it, the project usually cannot clearly explain what it does.
Unified risk model
Run every protocol through these ten categories. In real incidents, they often appear together: oracle manipulation triggers cascading liquidations, which drain liquidity, leaving collateral impossible to sell.
| Smart Contract Risk | Has the code been audited? Who controls upgrades? |
| Liquidity Risk | When you want to exit, will there still be money in the pool? |
| Oracle Risk | Where does the price come from? Can it be manipulated? |
| Bridge Risk | What actually backs the asset that crossed over? |
| Custody Risk | Who holds the assets? Who holds the private keys? |
| Counterparty Risk | If your counterparty disappears, what do you still have? |
| Governance Risk | Can a single vote move the money away? |
| Leverage Risk | If prices fall 30%, how many cascading liquidations follow? |
| Token Risk | Who gets diluted by unlocks and new issuance? |
| Operational Risk | How likely are errors in multisig, key management, and operations? |
See chapter 24 for instructions. Structure the risk section of a research report around these ten items, and explicitly mark anything you cannot assess as “Not Assessed.”
A complete example: a new high-yield pool
Use all three tools together.
Begin with questions 5 and 6: who pays, and where does revenue come from?
Open the documentation and identify three possible yield sources: trading fees, lending interest, and token subsidies. Calculate each share. If subsidies account for 80%, the rest of the analysis must revolve around when they stop.
Then ask question 8: who supplies liquidity?
Check pool depth and the distribution of LP addresses. If most liquidity comes from the project itself or two or three addresses, “you can exit at any time” does not hold.
Run through the risk model.
Focus on four items: whether contracts are audited, how the oracle obtains prices, who can upgrade the system, and the liquidation parameters. If the documentation does not answer them, the answer is “unknown risk,” not “no risk.”
Finally, ask questions 11 and 12.
Assume the token price falls 80% and all subsidies stop. What remains of the pool? If the answer is “nothing,” it is a subsidy distribution program, not a yield product.
These four steps take about 40 minutes. They cannot guarantee a profit, but they clarify what risk you are taking, and why.
How AI can speed up this process
AI can extract documentation fields, build comparison tables, list facts to verify, and generate monitoring rules.
You must verify data, judge source reliability, form the final conclusion, and bear the consequences yourself.