Write reasoning as numbered steps with explicit warrants — mark transitions that rely on assumptions
Externalize reasoning chains by writing numbered steps where each step connects to the next through an explicit warrant stating why step N leads to step N+1, marking any transition that relies on unstated assumptions.
Why This Is a Rule
Internal reasoning feels like a chain of connected steps, but the connections between steps are often assumed rather than verified. "A, therefore B, therefore C" feels logical, but does B actually follow from A? Is there an unstated assumption bridging them? Internal reasoning skips these checks because working memory treats the felt connection as a verified one.
Writing the reasoning as numbered steps with explicit warrants forces each connection to be stated and evaluated. "Step 1: The API is slow. Step 2: Therefore, we need a caching layer. Warrant: this assumes the slowness is due to repeated identical queries rather than expensive single queries." The warrant makes the assumption visible — and once visible, it's testable. If the slowness is actually from expensive single queries, caching doesn't help, and the reasoning chain breaks at step 2.
The primary diagnostic output is the set of marked transitions — the steps where the warrant relies on unstated assumptions, unverified facts, or feelings rather than evidence. These marks are your reasoning's weakest links, and they're exactly where investigation should focus.
When This Fires
- Before committing to a multi-step plan or argument
- When building a case for a decision, proposal, or recommendation
- During any complex reasoning where the conclusion has significant consequences
- When your reasoning "feels right" but you want to verify its logical structure
Common Failure Mode
Writing numbered steps without warrants: "1. Market is growing. 2. We should expand. 3. We need more engineers." Each step sounds reasonable, but the transitions are unsupported. Does market growth warrant expansion? (Maybe not — it depends on whether growth is in your segment.) Does expansion require more engineers? (Maybe not — it might require better tools or different partnerships.) Without warrants, the chain looks solid while being structurally fragile.
The Protocol
When externalizing reasoning: (1) Write each step as a numbered claim. (2) Between each pair of steps, write a warrant: "Step N leads to Step N+1 because [explicit reason]." (3) For each warrant, assess: is this based on evidence, assumption, or feeling? (4) Mark transitions where the warrant is assumption-based: [ASSUMPTION]. (5) The [ASSUMPTION] marks are your investigation targets — verify or replace these before treating the reasoning chain as sound.