Wiki-Grounded Reflection Loop
Journal freely, then connect current struggles to knowledge saved over time
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 98%
The loop starts with an unstructured end-of-day brain dump covering accomplishments, things built, conversations, unresolved struggles, and whatever remains mentally active. Instead of leaving the entry as a static diary, an LLM reads it and searches the personal wiki for relevant concepts, strategies, and source material. The response is grounded in those saved sources first, then supplemented with clearly distinguishable general model knowledge. This creates advice tailored not only to the immediate journal entry but also to the user's chosen intellectual inputs over time. The user reviews the connections, rejects weak or irrelevant matches, and selects a concrete next step. Repeating the process produces a searchable record of recurring concerns, applied ideas, and changes in judgment.
Origin
Extracted from Marketing Against The Grain from Matt Wolf's practice of journaling directly into Codex and cross-referencing each entry with his Obsidian wiki.
Core principles
- 01Unstructured reflection reveals concerns that formal prompts miss
- 02Current problems become more tractable when linked to prior learning
- 03Personal sources should ground advice before general model knowledge
- 04Daily accomplishment and struggle both deserve review
- 05Useful reflection ends with relevant options or next actions
How to run it
- 1
Capture the day
Record accomplishments, work completed, conversations, and events that felt important. Use plain language rather than trying to create a polished account.
Pro tip Journal at a consistent time near the end of the day.
Watch out A performative journal can conceal the issues that most need reflection.
- 2
Name unresolved friction
Write down struggles, missing ideas, uncertain decisions, or thoughts that continue to occupy attention. Do not require a solution before recording them.
Pro tip Include the next deadline or context surrounding each unresolved issue.
Watch out Do not outsource high-stakes judgment to the model.
- 3
Brain-dump freely
Add anything else on your mind without classifying it. The purpose is to expose connections that a rigid form might exclude.
Pro tip Keep capture separate from analysis so self-editing does not interrupt reflection.
Watch out Review the entry before exposing highly sensitive material to any external model.
- 4
Ground against the wiki
Ask the model to identify saved sources and ideas relevant to the journal entry. Require it to show which recommendations come from the wiki.
Pro tip Tell the model to prioritize personal sources before its general training knowledge.
Watch out A retrieved source may be relevant in topic but wrong for the current situation.
- 5
Add general knowledge carefully
Allow the model to supplement wiki-grounded insights with broader knowledge while clearly distinguishing the two. Treat both as suggestions rather than authority.
Pro tip Ask for tensions or disagreements between saved sources, not just confirming advice.
Watch out The model may blur retrieved evidence with generated interpretation unless instructed otherwise.
- 6
Commit to a next step
Choose one practical action, experiment, or question arising from the reflection. Record it with the journal entry so the next review can assess what happened.
Pro tip Favor a small reversible experiment when uncertainty remains high.
Watch out Reflection without action can become another form of avoidance.
In the wild
Matt journals that he needs a Tuesday video idea but does not yet have one. Codex cross-references the entry with mindset and YouTube growth material in his wiki, then combines those sources with broader model knowledge to suggest relevant approaches.
→ A current creative block is addressed using personally curated material rather than generic brainstorming alone.
At the end of the day, a founder records completed work, important conversations, and a persistent decision problem. The agent retrieves related notes and frameworks from the founder's wiki and presents a few grounded options for the following day.
→ Daily reflection produces continuity between accumulated learning and tomorrow's priorities.
Common mistakes
Polishing instead of brain-dumping
Trying to write an elegant journal can filter out uncertain or uncomfortable thoughts. Capture honestly before asking the model to structure the material.
Accepting superficial source matches
Topical similarity does not guarantee that a saved idea applies to the current problem. Inspect the retrieved source and its original context.
Letting the model become the authority
Personalization can make generated advice feel unusually convincing. Treat it as reflection support, not professional judgment or a replacement for agency.
Is it for you?
Best for
It is best for journalers, creators, and founders who want recurring problems connected to relevant ideas they previously saved.
Not ideal for
It is not ideal as a substitute for professional mental-health, medical, legal, or financial guidance.
From the transcript
“So at the end of every single day, I love to just sort of journal on my day.”
“And what it actually does is it cross-references my journal with everything that's in my wiki.”
“it will sort of ground itself with the wiki first”
From the episode
This AI Second Brain Remembers Everything I Save (Codex)