thala-dev
A place to share development practices that emerge from building thala.
About thala
Thala might end up being a personal knowledge system that goes beyond storage, to think alongside you. For a decade I ran this as a done-for-you service: acting as a “second brain” for clients, knowing their thought processes and beliefs well enough to have a “mini-client” running in my subconscious. This project attempts to automate that relationship, while throwing off some new research material for me to read.
The goal is to give technical users their own second brain—with background processes that mirror your own default mode network. It’ll maintain coherence, recognizes patterns, and develops genuine context with you over time.
The project is in early development. Currently building out the research workflow layer—the “executive” functions that gather and synthesize information. The more difficult parts (coherence layer, background processing, the actual “second brain” behavior) come later.
Development workflow
Thala has so far been built mostly using Claude Code with something I’ve adapted from Every’s Compound Engineering workflow. The core loop is the same as theirs:
flowchart LR Plan --> Work --> Review --> Compound -.-> Plan
Each cycle compounds: plans inform future plans, reviews catch more issues, patterns get documented. The agents in thala’s .claude/agents/ directory are mostly spun off from Every’s compound-engineering-plugin, customized to my project’s needs.
Every’s documentation stack is excellent. I’ve extended it with a /worth-sharing workflow that looks at a pattern or solution and decides whether it’s worth sharing publicly. If it is, both a GitHub gist and an article for this site are created. The aim being to make patterns and solutions more easily searchable as individual entities - so that when your agent looks for something my agent has solved, it’s easy to find.
What’s here
This site collects development practices worth sharing—patterns, solutions, and workflows that emerged from building thala. And all I do is say “is that worth sharing?”, or add /worth-sharing to a todo routine.
Posts so far: