Marco Arment's New App: A Verify-Before-You-Commit Guide to Capturing Tacit Knowledge

This guide applies a verify-before-you-commit rule to Marco Arment’s New App.

It focuses on using its micro-prompts to capture tacit knowledge, with the under-90-second threshold and verification gate detailed in the How It Works section below.

How It Works

Marco Arment’s new app uses a short capture loop: one micro-prompt surfaces a single question, you answer from memory or recent practice, the app structures that answer as knowledge, and the flow stops at a verification gate before anything is committed. The guide’s target is to keep the whole pass under 90 seconds by limiting scope to one prompt at a time.

A micro-prompt is a brief, specific question meant to draw out tacit knowledge—the practical know-how that is hard to state directly, as AFFiNE’s knowledge-management overview frames the explicit-versus-tacit distinction. A live complete option is the choice you are about to accept with its current availability and full terms visible, not a cached, partial, or assumed version. Commit means the final act of saving or sharing that captured item. Like-for-like means comparing options on the same basis, including totals and terms, rather than mixing headline figures with different conditions.

The verification gate is the reader rule in action: before commit, confirm the option is live and complete, then compare like-for-like totals and terms. The mechanism does not treat a fast answer as permission to skip that check; speed belongs to capture, not to acceptance.

In practice, a prompt should be answerable without research. If you need to look up a policy, price, or exception, the prompt is probably asking for explicit knowledge rather than tacit knowledge, and the better move is to defer it instead of forcing a guess into the 90-second window.

This section stays with mechanism and definitions. Decision criteria, pitfalls, non-obvious tactics, and side-by-side comparisons are handled in their own sections so the capture loop remains the single focus here.

Key Factors to Consider

This section alone lists the top decision criteria and numbers. The first criterion to verify is distinctness. Before committing any captured answer, check if it is tacit knowledge that is not already recorded elsewhere. The 2012 book by John Yearwood and Andrew Stranieri, Approaches for Community Decision Making and Collective Reasoning: Knowledge Technology Support (IGI Global), is cited here as background on structured group reasoning; the sources do not verify any specific capture rule from it, so treat the distinctness test as this guide's own check: search the existing record first, and commit only if the item adds a judgment, exception, or sequence the record does not already contain.

The second criterion is verification load. The framing that knowledge expands possibilities while a decision closes them appears in Beyond Knowledge; the sources list Sowell's Knowledge And Decisions only as a related title without verifying any specific claim from it, so do not attribute the expansion/closure line to Sowell. The check itself stands: can the verification gate be cleared using only the answer you just gave, or would committing require pulling in outside context to be sure? If outside context is required to satisfy that gate, the item is not ready to commit under a verify-before-you-commit rule.

The third criterion is reuse potential. From the framing of Beyond Knowledge, a decision defines what survives. For this reason, do not commit an item without identifying at least one specific decision it would inform later. If no concrete use can be named at the moment of verification, the act of committing cannot be justified by the same like-for-like standard used for items that do pass this check.

As for numbers that matter, the available sources provided does not supply any verified price, percent, fee, or count tied to this capture flow. Because of that, the only valid approach is method-only: for each candidate item, run the three criteria against one another and compare like-for-like totals of items that pass all three, not totals of items that were merely captured. This keeps the decision to the terms you can verify, without relying on any figure that is not present in the sources.

Common Mistakes

The first common mistake is committing a captured answer that is already explicit knowledge. A note such as “restart the service after every configuration change” may feel like hard-won judgment, but if the runbook, ticket, or wiki already states it, the commit adds a duplicate that competes with the authoritative source during later retrieval. Before committing, search the live, complete record across documentation, tickets, and prior captures, and commit only when the answer adds a missing judgment, exception, or sequence that the record does not already contain.

The second mistake is treating a memory as complete when it is only a fragment. An answer drawn from one incident can omit the condition that made the action work: “ship on Friday because the deploy window is quiet” may depend on one team’s traffic pattern rather than a general rule. Before committing, compare the captured answer with the full context of the original situation, including who was involved, what changed, and what would have to be true for the same action to be safe elsewhere. If those terms do not match, keep the answer scoped to its original conditions or discard it.

A useful check is to verify the live, complete option before committing and to compare like-for-like terms. That means the captured statement, its source context, and any competing record should be reviewed together rather than trusting the most recent or most confident version. When the terms differ, the safer choice is to store the narrower claim, add the missing condition, or leave it uncommitted.

Speed is not the same as quality. A fast capture that skips this review only stores noise more quickly and makes later decisions harder. Judge success by whether someone who did not share the original moment can find the item, understand its limits, and reuse it without needing the author present. That standard matters more than how quickly the answer was captured.

Insider Tactics

The non-obvious strategy is to treat a micro-prompt answer as a hypothesis until the verification gate, not as a finished note. Answer from recent practice, then ask whether the answer would change a decision or merely restate a procedure you already follow. If it only restates, leave it uncommitted; that keeps the captured set focused on knowledge that is actually tacit.

The timing tip is to separate capture from commitment. Capture while the task cues are still fresh, then return to the verification gate after a deliberate pause and ask whether you would still endorse the answer now. This prevents a fluent first response from being mistaken for a durable rule.

Before committing, attach the trigger that made the answer useful. A tactic without its condition is hard to reuse, so record the situation, constraint, or failure mode alongside the answer. At the gate, check that the trigger is specific enough that another person could recognize when to apply it.

When two captured answers compete, compare like-for-like totals and terms before choosing one. Normalize scope, assumed environment, and level of detail; an entry that includes setup steps should not be judged against one that assumes them. The better entry is the one that remains useful after both are stated on the same basis.

Use the gate for one disconfirming check: name the boundary or case where the answer would fail. If you cannot state that boundary, the answer is probably too broad to commit and should be split or discarded.

The micro-prompt path wins when the answer is tacit, specific, and not already documented. Its gate is the deciding advantage because it prevents a duplicate from becoming committed knowledge. The open-note path wins when you need a quick record and can tolerate later cleanup, but it loses on distinctness because nothing stops an already-explicit answer from being saved again. The skip path wins when the answer is already explicit knowledge or when the prompt surfaces nothing beyond what the record already contains.

Before choosing, verify the live, complete option: confirm the prompt, answer, and gate are all present in the current flow, not a cached or partial version. Do not compare totals by capture speed alone. A faster commit that adds duplicate knowledge is not a better total; compare the committed result, not the seconds spent. If two options produce the same committed knowledge, choose the one with fewer verification steps only after confirming the distinctness check has already happened elsewhere.

Winner: use the micro-prompt path with the verification gate for distinct tacit knowledge; use skip for already-explicit knowledge; reserve open notes for cases where the gate is unavailable and you will perform the distinctness check manually before commit.

What to do next

Frequently Asked Questions

What is the target time for one full capture pass with Marco Arment’s new app?

The guide’s target is to keep the whole pass under 90 seconds by limiting scope to one prompt at a time.

At what point in the capture loop is commitment prevented?

The flow stops at a verification gate before anything is committed.

Which 2012 publication is cited for approaches to community decision making and collective reasoning?

The cited work is Approaches for Community Decision Making and Collective Reasoning: Knowledge Technology Support by John Yearwood and Andrew Stranieri, published by IGI Global in 2012.

What type of know-how is a micro-prompt meant to draw out?

A micro-prompt is meant to draw out tacit knowledge—the practical know-how that is hard to state directly.

What is the tradeoff between knowledge and a decision?

Knowledge expands possibilities, while decision closes them.

What is defined as a 'live complete option'?

A live complete option is the choice you are about to accept with its current availabil

Quick answers

Also worth reading: Sign AI Training Terms at Capture: Where C2PA Binding Breaks: Sign AI Training Terms at · New Manager Coaching Tools: 112 vs 140 Days to Competency: New Manager Coaching Tools: 112 · New hire coaching: Retrieval-Augmented Coaching (RAC) cuts 31% vs mentor 2026: New hire coaching: Retrieval-Augmented Coaching