The quality of a PKM system is bounded by the quality of what enters it. No amount of clever organization, linking, or retrieval can compensate for a diet of low-quality, ephemeral, or irrelevant information. Curating your information inputs — deciding what you consume and what you ignore — is the upstream problem that determines downstream PKM health. Garbage in, garbage out applies to knowledge management as much as to computing.
The Information Diet Concept
Clay Johnson coined the term "information diet" in his 2012 book of the same name, drawing an analogy to food: just as a diet of junk food degrades physical health, a diet of junk information degrades cognitive health. The analogy is imperfect but useful. Most knowledge workers consume far more information than they can process, much of it optimized for engagement rather than insight.
An information diet is a deliberate set of choices about what sources you subscribe to, how much time you allocate to consumption, and what filtering criteria you apply. It is not about consuming less in absolute terms but about consuming better.
Filtering vs. Hoarding
A common PKM anti-pattern is hoarding: saving everything "just in case." This creates a growing backlog of unprocessed captures that eventually becomes psychologically overwhelming and practically useless. See PKM Anti-Patterns for more on the collector's fallacy.
Effective filtering happens at multiple levels:
- Source level: Unsubscribe from low-signal sources. Ruthlessly prune RSS feeds, newsletters, and social follows. The goal is a small number of high-quality inputs, not comprehensive coverage.
- Capture level: Not everything worth reading is worth saving. Apply a "will I use this?" test before capturing. If you cannot articulate a specific use case, do not capture it.
- Processing level: Captured items get a triage pass. Most will be discarded after a quick review. Only the genuinely valuable items proceed to permanent notes.
This three-stage filter is similar to Progressive Summarization, which applies layered highlighting to surface the most important material.
Read-Later Workflows
Read-later tools (Readwise Reader, Omnivore, Pocket, Instapaper) serve as a buffer between discovery and processing. They prevent the "I need to read this now" impulse from hijacking deep work. But they also create their own failure mode: an ever-growing read-later queue that becomes another source of guilt.
The discipline is to treat the read-later queue as a FIFO buffer with a maximum size, not an infinite archive. If you have not read something within two weeks, it probably was not that important. Periodic purges of the read-later queue are essential. Readwise's integration with Obsidian and other PKM tools makes it particularly useful for moving highlights into a knowledge base, but only if the upstream filtering is sound.
The Lindy Effect as an Input Filter
The Lindy Effect, popularized by Nassim Taleb, states that the longer something has survived, the longer it is likely to continue surviving. Applied to information inputs: a book that has been in print for 50 years is more likely to contain durable insights than a blog post published yesterday.
This does not mean ignoring new information. It means weighting your input mix toward time-tested sources. Read classic books, foundational papers, and long-running publications. Use recent sources primarily for developments in fast-moving fields (technology, current events) where recency genuinely matters.
Feynman's Twelve Favorite Problems
Richard Feynman reportedly maintained a list of roughly twelve open problems that he cared about. Whenever he encountered a new piece of information, he tested it against his list: does this help me make progress on any of my favorite problems? If not, he moved on.
This is one of the most powerful input filters available. Maintaining a short list of questions, themes, or projects that matter to you creates a clear decision criterion for what to capture and what to ignore. It transforms consumption from passive browsing into active research. This practice connects naturally to the goals and projects that drive Building a Second Brain's PARA method.
Perishable vs. Timeless Information
Not all information has the same shelf life. News is perishable by definition; it loses value within days or hours. Technical documentation is semi-perishable; it is valid until the next version. Principles, mental models, and well-established science are timeless; they remain useful for years or decades.
A well-curated information diet tilts heavily toward timeless information. The ratio matters: if 80% of what you consume is perishable, 80% of your PKM effort is wasted on notes that will be irrelevant within months. This connects directly to Knowledge Lifecycle and the problem of knowledge rot.
AI-Powered Feed Curation
AI tools are increasingly capable of helping with input curation. LLM-powered summarizers can pre-digest long articles so you can decide whether to read in full. Recommendation algorithms (when tuned for quality rather than engagement) can surface relevant material from large corpora. AI agents embedded in PKM workflows can flag incoming items that relate to your existing notes or active projects.
However, AI curation carries its own risks: filter bubbles, over-reliance on algorithmic judgment, and the temptation to automate away the sense-making that is the most valuable part of PKM. The goal is AI-assisted filtering, not AI-replaced thinking. See AI Skills in PKM for broader coverage of AI integration points.
Key Points
- PKM quality is bounded by input quality; no system can compensate for a poor information diet
- Filter at three levels: source, capture, and processing
- Read-later queues need size limits and periodic purges to avoid becoming guilt-inducing backlogs
- The Lindy Effect suggests weighting inputs toward time-tested sources
- Feynman's "twelve favorite problems" is a powerful filter: does this input advance a question you care about?
- Distinguish perishable from timeless information and tilt your diet toward the latter
- AI can assist with curation but should not replace the sense-making that gives PKM its value
Open Questions
- How do you audit your information diet systematically? What metrics indicate a healthy vs. unhealthy mix?
- Can AI curation avoid creating filter bubbles that narrow your intellectual exposure?
- What is the right balance between serendipitous discovery and focused, goal-driven consumption?
References
- Clay Johnson, "The Information Diet" (2012)
- Nassim Nicholas Taleb, "Antifragile" (2012) — on the Lindy Effect
- Richard Feynman, as relayed by Gian-Carlo Rota, "Ten Lessons I Wish I Had Been Taught" (1996)
- Tiago Forte, "Building a Second Brain" (2022) — PARA method and capture criteria