Best Practices

Start With A Question

Before adding sources, decide what kind of signal you want:

• reproducible tactics;
• non-obvious ideas;
• cases and decisions;
• tool or source references;
• material for notes, content, or research.

If you save everything into a folder, vault, or RAG system, future search often returns many related but weak fragments. Knowlume is most useful before that step: it helps you separate candidates worth saving from background noise.

Select Before You Export

Do not export a whole source by default. After processing:

1. Open the fragments.
2. Filter by reproducibility, originality, and sourcesness.
3. Inspect the strongest candidates.
4. Save only what will be useful later.
5. Export the saved selection.

This keeps your knowledge base compact. It stores fragments you can inspect, apply, or reuse, not every generated summary.

Batch Your Inputs

For a recurring stream of material, adding sources one by one is slow.

Practical options:

• send links and posts through @knowlume_bot;
• use the mobile share action;
• forward useful Telegram posts to the bot;
• copy a list of open tabs with Tabcopy or a similar tool;
• add multiple videos, articles, or pages into one project.

After batching, do not try to read every result immediately. Wait for processing, then scan the strongest fragments inside the project.

Split Projects By Job

A project should match a real boundary: conference, course, research topic, client, product area, or personal inbox.

Good examples:

• "AI development conference";
• "SaaS growth cases";
• "saved Telegram posts";
• "knowledge base tools";
• "article research".

If every topic goes into one project, the filters still work, but selection and export become harder.

Do Not Trust Prebuilt Guides Blindly

Curated guides, playlists, and AI summaries can look useful while hiding several problems:

• some items may contain no useful knowledge;
• ideas may repeat across sources;
• a strong case may sit inside a talk with an unhelpful title;
• a generic summary may list ideas that are not actually supported by the source;
• without fragments and context, every recommendation must be validated manually.

It is usually better to process the original material, select fragments, and keep only the cases that survive inspection.

Treat Conferences As Research

For conferences, courses, and large YouTube playlists, create a dedicated project.

Workflow:

1. Add the relevant videos or materials.
2. Wait for processing.
3. Filter fragments by reproducibility and originality.
4. Review strong cases in a focused session.
5. Save only fragments you can apply or use in a model.
6. Export the result to Obsidian, trip2g, or another note system.

For case libraries, a simple structure often works well: problem, lever, target state. That shape is easier to reuse than a long talk summary.

Use Knowlume For Saved Items

Saved posts, chat links, and browser tabs quickly become an archive nobody revisits.

Instead:

• send posts and links to Knowlume as they accumulate;
• keep a separate project for them;
• review strong fragments weekly;
• leave weak sources unexported;
• move useful fragments into your working knowledge base.

This makes Knowlume a filter before your knowledge base, not another infinite collection.

Automate After The Manual Flow Works

Webhooks and the public interface are useful after you know what result you want manually.

Good automation scenarios:

• submit sources from RSS or another service;
• send transcripts from recorders or transcription tools;
• deliver selected fragments into a note system;
• post-process exports for a content pipeline;
• classify fragments and route them inside a knowledge base.

Do not start with automation. First tune the manual selection flow on a few sources, then automate the repeated part.

Verify Context

Knowlume helps find knowledge candidates, but it does not remove judgment.

Before saving an important fragment:

• open its context;
• check that the conclusion is supported by the source;
• do not save generic advice just because it is well-written;
• do not use a fragment as evidence if it lacks a reference, quote, or enough context;
• use domain expertise for legal, medical, and financial decisions.

A good knowledge base is built from a small number of trusted, reusable fragments, not from the maximum possible number of notes.