Journal

What’s On My Mind

What does strong schema formation look like across skill-based vs concept-heavy domains — and at what point is a schema detailed enough to generate novel applications reliably?
How should I use the knowledge base as a thinking partner instead of only an archive?
How do I use the knowledge base to free attention from bookkeeping without weakening my own thinking?
Is it possible to compress the learning system into a minimally viable system?
Can that compression also fit inside the five learning dimensions, so each dimension can be improved separately?
What usable frames should guide each Priority 0 skill: Agentic Engineering, Learning Systems, Vietnamese, 中文, Fitness, and Relationships?

Current Focus

Are You Learning, or Just Using Techniques: A technique only matters if it triggers the thinking the result requires. Beware clean-looking activity that avoids comparison, reconstruction, uncertainty, or judgment.
First Principles of ICS: Overwhelm is disorganized information. Turn confusion into a question, then use the question to organize the next move.
ICS System: Control the learning process: prestudy, encode deeply, retrieve intelligently, reflect, and upgrade the system one constraint at a time.
Prestudy, BHS, and SIR: Turning Information into Usable Structure: Prestudy gives the frame, BHS builds the map, SIR keeps it retrievable. The goal is usable structure: building a stable mental model where information can sit.
Balancing Multiple Interests: Breadth v Focus: Keep the full identity alive without making every interest active at once. Rotate deliberately, preserve momentum, and avoid false abandonment. Breadth belongs in Priority 0. Focus belongs in the active loop.
How to Unlearn Old or Bad Habits Efficiently: Old habits win because they’re cheap and automatic. New skills lose because they’re expensive. Isolate the cue → old response, write a replacement script, rehearse judgment until the new move feels obvious. Practice the bottleneck.
Minimally Viable Learning System: The current minimum is BHS + SIR: encode usable structure, then retrieve, space, interleave, and repair it. Agentic Engineering and the knowledge base are adjacent systems being tested, not part of the minimum yet.
Minimalism as Systems Design: Reduce ownership cost so the environment becomes easier to act inside. Empty space is reserve capacity. Every object, workflow, subscription, and commitment charges attention, maintenance, and optionality.

Active Threads

ICS System The learning system that anchors the study notes. Bear Hunter System Encoding workflow built around Aim, Shoot, and Skin. Spaced Interleaved Retrieval Retrieval system for spacing, interleaving, and gap repair. Attention Management Managing attention across tasks without breaking flow. Agentic Engineering Best practices for building with LLM agents. Red Teaming Decision support, assumptions, groupthink, and cultural empathy.

Possible Blog Posts

Emacs to Obsidian: Emacs was more fun to use because it ran on macros and keybind muscle memory. Moving around Emacs felt closer to playing World of Warcraft than using a normal app. Obsidian, by comparison, feels more like playing with a mouse only. People known as “clickers” have a bad reputation. But it’s not entirely a fair assessment due to different UI styles. I definitely feel slower in Obsidian.
How I Built This: a dated account of building this site from Karpathy’s LLM knowledge-base idea, Obsidian, Quartz, and several AI coding agents as collaborators.
On Red Team’s Closure: draft commentary arguing that Red Team’s closure was understandable because the institutional version had become too bloated and time-inefficient, while the lean two-week course was still worth preserving.

Private Workflow Templates

These are blank public shells for private Obsidian workflows. The templates can be copied into a private vault, but filled-in goals, skills, and Kolb’s reflections should stay local.

Recent Log

2026-05-17: Expanded the agentic side of the knowledge base. Added Current Agentic LLM Stack, refined Hybrid Model Workflows, Grok + Hermes, and updated AGENTS with model-specific guidance for Grok vs Hermes, local-agent access boundaries, and the raw/sessions/ session-summary habit. The operating question is shifting from “which model is best?” to “which model should own which layer of the workflow?”
2026-05-17: Added the 30-Day Challenges section across the five learning dimensions: Mindset, Self-Management, Self-Regulation, Deep Processing, and Retrieval. The challenges make each dimension practiceable instead of purely descriptive: four weekly arcs that turn the dimension model into concrete experiments.
2026-05-17: Continued refining How Top Performers Learn around learning assets and the Five Dimensions. The page now connects technique mastery to four requirements: understand, remember, apply, and perform consistently under pressure.
2026-05-16: Refined the post-restructure wiki around learning-system architecture. The five learning dimensions now read more like diagnostic control surfaces, with clearer failure modes and supporting techniques for Deep Processing, Retrieval, Mindset, Self-Management, and Self-Regulation. Added How Top Performers Learn as the bridge between top-performer learning, learning assets, and the Five Dimensions.
2026-05-16: Added Hybrid Model Workflows, Grok + Hermes to document the current multi-model workflow: Grok for high-level framing and coherence, Hermes/Aider for local file execution, link hygiene, and structural cleanup. This makes the agent workflow itself part of the knowledge-base architecture instead of just an invisible work habit.
2026-05-16: Restructured the knowledge base around the actual operating model instead of the old generic buckets. Study pages moved out of the flat Techniques/Concepts sprawl and into Dimensions of Learning: Deep Processing, Retrieval, Mindset, Self-Management, and Self-Regulation. AI and tooling pages moved into Systems, while the public navigation stayed cleaner and less overloaded. This makes the site feel more like a working system: dimensions as control surfaces, systems as domains of practice, and outputs as the staging layer before durable wiki integration.
2026-05-16: Created GPT - How to Rest So Well You Never Feel Exhausted Again and moved the remaining clipping inbox files into local raw sources. The useful recovery frame: rest has to match the depleted system. Low-effort stimulation can keep attention active, while real recovery uses detachment, mastery, control, and low-load environments to restore capacity.
2026-05-15: Created output briefs for Live Clinic 61: Opening the Black Box of Learning and Watch This To Force Your Brain To Study FASTER. Live Clinic 61 reframes the black box as the inspectable process layer between learning inputs and outputs. Study Faster sharpens the schema formation model: construction, assimilation, and frequent reorganization.
2026-05-14: Created Schema as a dedicated concept hub. Three cognitive pillars (construction, assimilation, reorganization) map directly onto Prestudy/Aim, Shoot, Skin. Logged the schema assimilation vs accommodation distinction in the calendar — assimilation extends the existing schema; accommodation restructures it when the shape is wrong.
2026-05-14: Added How to shift your brain to be motivated — DFUZ and thought-action defusion. Motivation dependence makes consistency structurally impossible; the fix is separating feeling from action, then using the bidirectional loop (actions change thoughts change feelings) to build uptime gradually.
2026-05-14: Compiled wiki pages from Live Clinics 64, 65, and 66 — Deep Processing for Research, How to maintain sustainable energy under pressure, and How to prepare for ultra high-volume exams. Added Best-attempt Encoding as a standalone practical guide.
2026-05-13: Reflected on Knowledge Base as Thinking Partner and the risk that the system could create passive connection-browsing instead of stronger first-brain schema building.
2026-05-13: Added Minimalism as Systems Design to Current Focus and set Minimalism graph nodes to black in light mode and white in dark mode.
2026-05-12: Simplified Home into a graph-first entrance and moved the longer project explanation into About.
2026-05-12: Added Current Focus cues to the Journal and Home pages, compiled How to Unlearn Old or Bad Habits Efficiently, and folded the older unlearning page into the new synthesis.
2026-05-11: Built the first large outputs layer from GPT, Grok, and Opus synthesis passes, then converted the highest-value Live Clinic work into ICS synthesis pages.
2026-05-10: Made raw sources local-only, cleaned the wiki architecture, regenerated Applied Critical Thinking, and started the outputs/briefs workflow at scale.
2026-05-09: Pruned the Red Team section down to the identity page, mapped the deleted wiki pages to what I would cut from the handbook, and drafted commentary on Red Team’s closure.
2026-05-09: Added Journal as the public thinking dashboard for current questions, active threads, possible essays, and a light dated log.
2026-05-08: Tightened the public site around the knowledge base: homepage, graph behavior, clean layout, Index, and active learning/decision-making pages.
2026-05-07: Converted the project into a personal Quartz site, added the Red Team and ICS hubs, and iterated on public navigation, graph color, typography, and visual style.
2026-05-06: Created the initial Quartz-backed repository and published the first version of the LLM-maintained knowledge base.

Rituals / Review Prompts

What question keeps returning?
What active thread deserves one focused session?
What belongs in the wiki, what belongs in a blog post, and what should stay private?
What did the knowledge base help me understand this week?
What page should be split, renamed, or retired?

Calendar

For the chronological view, see Journal Calendar.