6 Months of OpenClaw - NotesByLex.com
Recent<br>Tags
6 Months of OpenClaw<br>Jul 15, 2026
Completing A Computer Science Degree On Coursera<br>Jul 5, 2026
Shipping a Laptop to a Refugee Camp in Uganda<br>May 23, 2026
Heavy Thinking: A Test-Time Scaling Pattern for Hard Problems<br>May 17, 2026
LLMs Corrupt Your Documents When You Delegate<br>May 9, 2026
AI-Induced Cognitive Atrophy<br>May 7, 2026
OpenGame: Open Agentic Coding for Games<br>Apr 29, 2026
Naming Things Is Easy Now<br>Apr 25, 2026
Obsidian Markdown Notebook: code execution with outputs stored in the file<br>Apr 21, 2026
Research, Plan, Implement Workflow<br>Mar 22, 2026
See more →
MachineLearning<br>29
SoftwareEngineering<br>17
LinearAlgebra<br>16
LargeLanguageModels<br>13
GameDesign<br>12
AgenticReasoning<br>11
ComputerScience<br>11
AudioEngineering
LearningAndTeaching
Zettelkasten
DiscreteMath
AutomatedTesting
ReinforcementLearning
Roblox
DataStructures
Lua
StoryWriting
GameMath
Obsidian
LearningWithAI
See all 102 topics →
Home<br>/essay
6 Months of OpenClaw
Jul 15, 2026
OpenClaw<br>Obsidian
OpenClaw went viral earlier this year on tech internet, but the hype has died down quite a bit, especially if Google Trends is anything to go by.
But I'm still using it daily, and have been ever since I set it up in January, when it was called Clawdbot.
I use it as an entry point to my Markdown-based life admin and note-taking system. The combination of LLM agents, cron, and chat app connectors has proven very convenient.
I feel like it's the final piece to the promise of having a "second brain". I also love that I'm building an LLM Memory system that's vendor-agnostic; I can just switch to a different LLM (ideally, one day, a locally running one) and all my memories switch with me.
I wrote an article about my setup a few weeks into my OpenClaw journey, and my setup hasn't changed meaningfully since then (see OpenClaw: the missing piece for Obsidian's second brain). However, in this article, I want to reflect on what I actually achieved by running an OpenClaw instance and share a few things I've learned along the way.
My High-Level Workflow
Obsidian vault
I have a private Obsidian vault, which is a collection of Markdown notes and documents I've been accumulating for years. Every day, I create a daily note to plan my day. It also acts as a journal and a dumping ground for anything I need to remember.
Then, projects, people, trips and goals get separate notes, as does anything I want to track over time. Those notes can have their own journal, and I link the journal entries back to the daily note, so I can retrieve information by day or by topic.
I also keep permanent notes for ideas and reference notes for papers, books and courses. When a note becomes substantial enough, I'll move it to the public repo published on this site. On top of that, I maintain a centralised to-do list which tracks all my key tasks and deadlines.
Initially, my vault was maintained entirely by hand, but in recent years, it's become a hybrid of an LLM Wiki and a journal. I still do all writing by hand, but I'm also okay with an LLM managing and updating certain files in the vault, especially projects that are just information dumps.
OpenClaw and other agents
OpenClaw runs on an old laptop that I treat as a server. Its workspace is a folder in my vault, and OpenClaw is mostly instructed to refer to the parent folder's AGENTS.md file - which itself just explains how the vault is structured. This means that any other agents I want to work with, such as Claude Code or Codex, also have the same context as OpenClaw, and that I can use whichever tool I need for a particular job. I keep the shared skills in .agents/skills, which Codex and OpenClaw load directly and Claude Code accesses through symlinks.
I primarily use WhatsApp to communicate with OpenClaw. It's already the chat app that my family and friends use, and I bought a new SIM card to create a separate WhatsApp account for my OpenClaw agent, M .
Also, I do use Obsidian and an LLM wiki for work, but I keep it in a totally separate vault that doesn't sync with my personal vault and that OpenClaw never sees, per my work's security policy.
The main use cases
1. Calorie, weight and workout tracking
Last year, I set a goal of getting lean, and I have been trying to maintain a calorie deficit by meticulously tracking the calories in every meal I eat, much to the chagrin of those in my life. I've found LLMs to be a convenient way to do this, particularly because most of them can handle freeform information through text descriptions, images of nutrition labels, or images of the food itself.
OpenClaw consults a file called common-foods.md when I tell it what I've eaten. The file contains recorded calories and macros for foods I've eaten before. This is something I'm building over time to make the process more convenient - I can just tell it I'm eating my regular Subway order or that Musashi protein bar I usually buy, for example.
I'm aiming...