How Brali thinks

Try the move. Understand the mechanism. Know where the evidence stops.

Brali is a library of small protocols, not a shrine to productivity tricks. A useful page should tell you what to try, why the move is plausible, what signal to watch, and what would make you change or stop it.

1 · Do

Start with an action

If a page needs three paragraphs of terminology before you know what to do, the page has work to do. Brali starts with the smallest useful move.

2 · Understand

Learn the mechanism

The goal is not to borrow scientific vocabulary. It is to understand what the protocol is changing: attention, timing, friction, feedback, retrieval, environment, or another observable part of the situation.

3 · Observe

Watch a real signal

Track the outcome the protocol is actually meant to change. A streak is not automatically progress, and a timer is not a biological law just because it has a reassuring countdown.

4 · Bound

Keep the claim smaller than the hype

If a study supports recall, Brali should not quietly upgrade that to intelligence. If it supports an acute response, it does not become a longevity protocol on the walk from the abstract to the headline.

Why the trust labels exist

Search visibility and recommendation trust are deliberately separate. Brali keeps the historical library discoverable so old material can be audited and improved, but people and AI agents should not receive an unreviewed page as if it had already earned the same confidence as a checked protocol.

That is why a page can exist publicly and still be review-gated. "We have a URL" is not an evidence state.

The current quality model

The library currently contains 973 entries. All 822 public entries are included in the sitemap; 224 currently meet the trusted recommendation bar and 749 remain review-gated.

Reviewed · 46

Claims checked

Evidence-like claims have traceable sources where required, and the practical guidance has completed editorial review. Reviewed does not mean immortal truth; it means the current wording survived an actual source check.

Practical · 178

Low-risk practice

The entry is framed as practical guidance and does not contain evidence-like claims that require scientific support. Not every useful action needs a lab coat.

Pending review · 598

Searchable, not recommended

The public page is crawlable and evidence-labelled, but it stays outside normal trusted recommendations until review is complete. A citation-shaped object is not yet a reviewed source.

Restricted · 151

Higher review bar

Sensitive material remains visible for transparency and compatibility, with a clear safety boundary and no promotion into trusted recommendations.

What source review changes

Research rarely arrives as a neat commandment. A review may support one outcome but not another; a trial may work in one population, for one duration, with several bundled components. Brali records those boundaries because they change the protocol. The interesting question is not simply did the paper have a positive result? It is what action does this result actually justify?

When evidence is mixed, Brali can keep a bounded experiment, narrow an existing page, mark a source as a challenge, or decline to create a new hack at all. Finding a paper is the beginning of editorial work, not the ceremonial end of it.

Current claim debt

The claim gate inspected 973 records and found review markers in 786. 607 records currently carry unresolved claim debt. 0 unresolved claim-debt records are eligible for trusted recommendation.

Claim debt is a review queue, not a verdict that a statement is false. Enforced categories block normal trusted recommendation when review is incomplete; monitor-only categories remain visible so editors can inspect broad effect, mechanism, and research-language wording without pretending a regular expression can replace source review.

quantitative · enforced · 746 records. Percentages, explicit sample sizes, or participant counts that require precise source support.
effect-estimate · enforced · 59 records. Statistical effect estimates, intervals, or comparative magnitudes that require exact source and analysis boundaries.
first-party-result · enforced · 558 records. Claims about Brali, author, user, pilot, trial, experiment, or internal data results.
guarantee · enforced · 27 records. Guarantees, universal effectiveness, or unsupported proof language.
clinical-outcome · enforced · 26 records. High-confidence diagnosis, treatment, cure, prevention, symptom, or disease outcome language.
causal-assertion · enforced · 14 records. High-confidence causal proof, shown-effect, or cause language that requires a directly matching reviewed source.
causal-effect · monitor-only · 458 records. Broader causal or effect language that should be reviewed when used as a factual claim.
mechanism · monitor-only · 91 records. Biological or neurological mechanism language that can create unsupported scientific authority.
research-language · monitor-only · 717 records. Study and evidence language that requires traceable review before it supports a public claim.

Topic-level debt

Claim debt currently spans 34 mapped Topics. 177 debt records remain without a canonical Topic assignment. Topic counts are editorial workload signals, not evidence-strength scores.

Clear Communication · clear-communication · 99 debt records
Cognitive Biases · cognitive-biases · 58 debt records
Everyday Health Basics · health-basics · 42 debt records
Goals · goals · 32 debt records
Relationships · relationships · 24 debt records
Work Systems · work-systems · 23 debt records
Mood & Wellbeing · mood-wellbeing · 20 debt records
Skill Learning · skill-learning · 20 debt records
Task Initiation · task-initiation · 15 debt records
Language Learning · language-learning · 14 debt records
Idea Generation · idea-generation · 12 debt records
Critical Thinking · critical-thinking · 10 debt records

Open the public claim-debt report → · Machine-readable JSON

How an entry earns trusted recommendation

Start with a concrete action or protocol.
Detect evidence-like wording, quantitative claims, effect estimates, high-confidence causal assertions, first-party results, guarantees, clinical-outcome wording, and sensitive topics.
Record usable external sources when factual support is required.
Review exact wording and sources rather than treating a recorded link as proof.
Rewrite the action so the mechanism and evidence boundary are understandable to a non-specialist.
Keep every public entry discoverable, but include only reviewed and practical entries in the Trusted Protocol Feed and normal AI recommendations.

Inherited corpus corrections

The migrated corpus is preserved for provenance. When the same inherited claim appears across many entries, Brali can apply a reviewed editorial normalization before public pages and evidence states are generated. 72 reviewed normalization rules are currently registered, affecting 9 source entries in this build.

These corrections are published with their rationale, review date, reviewer, source, and application count rather than silently rewriting the historical corpus.

Health and mental health

Brali's Growth Library is general educational material and a starting point for personal reflection. It is not medical diagnosis or treatment guidance. Sensitive entries receive a higher review bar and stay outside normal trusted recommendations until reviewed.

Machine-readable transparency

Trusted Protocol Feed — reviewed and low-risk practical protocols.
Evidence status index — evidence state for the complete library.
Review queue — entries still awaiting editorial work.
Claim debt report — claim categories, enforcement state, Topic aggregation, debt reasons, source presence, and review metadata.
Indexing policy output — complete search visibility and the separate trusted-recommendation set.
Editorial normalization register — reviewed inherited-claim corrections and where they were applied.

The static build fails if a public hack is missing from the sitemap, still carries noindex, loses its evidence state, or leaks a review-gated entry into the Trusted Protocol Feed. Quality gates are deliberately annoying in the useful way: they stop a nicer sentence from quietly becoming a worse claim.

Use Brali like a small experiment, not a belief system.

Choose a relevant protocol, check its evidence state, try the smallest useful version in your context, observe the intended signal, and keep, change, or stop the practice based on what happens and what the evidence actually supports.

Explore the Growth Library