Evidence-first research checklist

EN source master · P01-L08 · 30 minutes estimated · needs_review

Read-only study. No wallet connection, real money, seed phrase, private key, signature or live trade. Duration is an editorial estimate; visual assets are specifications, not deployed components.

Research quality depends on keeping observation and interpretation separate. A checklist is useful only when it changes the wording of conclusions and prevents unsupported claims.

Define the claim before collecting material

Begin with one testable sentence. ‘This contract reports a particular supply at a particular block’ differs from ‘this project is safe’. The first has an identifiable record target; the second bundles many unknowns. Breaking a broad claim into fields makes missing evidence visible rather than hiding it under one confidence label.

Bind a source to its role

An API specification explains field meanings [RPC]; a token standard explains interface semantics [ERC20]. Neither documentation page proves that a fictional or historical balance existed. To support a value, point to the actual supplied record and its context. To explain a computation, state both the input records and the method.

A source-to-claim mapping should say which record field supports which statement and where the inference begins. Copying a URL to the end of a report is not enough when the URL only discusses a mechanism.

Apply the four states precisely

OBSERVED: a directly inspectable field or a publication's actual statement, described at that level. INFERRED: a calculation or interpretation with explicit assumptions. UNKNOWN: a needed fact is missing. INSUFFICIENT EVIDENCE: the available material cannot establish the proposed claim. These labels are not a ranking in which every unknown must eventually become positive.

The same source can support a narrow observation while leaving a broader claim unsupported. A token allowance is the amount a specified spender is permitted to spend on behalf of an account under the token contract's rules [ERC20]. A supplied allowance of zero describes one permission at one context; it does not establish the absence of every route to asset loss. A website's statement that a team controls an address is observed as a statement, while the underlying attribution may remain unsupported.

Choose a useful next request

Ask for a specific missing record, date or method. ‘More research’ is too vague. For a missing time, request the source context; for a supply discrepancy, request the exact contract/block and accounting definition. For an identity claim, do not request a seed phrase or use a funding link as a substitute for attribution.

Stop the report at its boundary

A concise conclusion can remain useful with unknowns: state the exact object, the measured facts, the calculation, and the claims not established. No financial action follows from completing this checklist. The purpose is a defensible research note.

• Claim: a specific proposition to test.
• Provenance: where a record came from and how it was obtained.
• Inference: conclusion derived under assumptions.
• Coverage: the part of the question the records address.

ILLUSTRATIVE dossier at SIM-BLOCK-8. R1: TOKEN-Q totalSupply raw=1000000, supplied decimals=2. R2: allowance ACCOUNT-A→SPENDER-B=0 for TOKEN-Q. R3: project page claims ‘Team controls A’. No independent identity record, reserve record or timestamp supplied.

Evidence panel with R1/R2/R3 linked to narrow claims and four labelled states. Alt: observed fields lead to a supply calculation; identity and safety remain unresolved.

The stipulated fields, transparent arithmetic and existence of the claim.

Comprehensive permissions coverage, human identity, reserve backing or guaranteed safety.

• Citing mechanism documentation as a record.
• Marking a broad claim OBSERVED because one field is visible.
• Replacing unknown identity with a cluster guess.

Write four claims covering the evidence states. Compute displayed supply, assess ‘all funds are safe’, and map each conclusion to R1/R2/R3 or a missing record. Propose two targeted follow-ups.

Submit a short evidence table and reasoning, using only the provided material. Suggested allocation: study 12 min, inspection/calculation 8 min, correction/quiz 10 min.

• One precise claim per row.
• Record pointer and context.
• Inference assumptions.
• Missing facts and unsupported conclusions.
• Targeted follow-up without secrets.

Research quality depends on keeping observation and interpretation separate. A checklist is useful only when it changes the wording of conclusions and prevents unsupported claims. The supplied material supports the stipulated fields, transparent arithmetic and existence of the claim. It does not establish comprehensive permissions coverage, human identity, reserve backing or guaranteed safety.

• Assign all four evidence states claim by claim.
• Write a bounded research conclusion.
• Request the smallest evidence that resolves a gap.

P02-L01 after reviewing this exercise and its prerequisites.

NONE. The authoritative catalog requires no product access for this lesson. Use the frozen/offline material.

• [RPC] Ethereum JSON-RPC — Receipt status, block context and network-scoped API results. Checked 2026-10-01.
• [ERC20] ERC-20 token standard — Token balances, optional decimals, transfer, approval and allowance interfaces. Checked 2026-10-01.

Synthetic values are author-created exercise inputs. Historical values, if any, must use the linked dataset and its narrower provenance. Source inspection is not independent editorial acceptance.