Open Research.
We believe in building in public. This means sharing our successes, but more importantly, documenting our failures. Below are the experimental projects we've undertaken and the hard lessons learned.
Thinking Patterns
Pivot / Failed ExperimentOur 1st attempt to automate strategic AI reasoning frameworks using Ollama and Claude Code.
The Hypothesis (1st Attempt)
We tried using Ollama with Claude Code to design thinking patterns that would log every single thinking step into a Markdown (.md) file specifically within the `thinking-patterns` directory. The goal was that every time we ran a process via chat, it should generate and structure its reasoning process into these files.
The Failure
Our 1st attempt failed in just 1 day. The AI models failed to consistently manage the depth of logic required to log every thinking step accurately via chat, yielding inconsistent and non-deterministic logic structures instead of the seamless Markdown logs we needed.
The Pivot: 2nd Attempt
We are now initiating our 2nd attempt. We are pursuing the exact same true goal: using Ollama and Claude Code via chat to log every thinking step in a Markdown file inside the `thinking-patterns` directory every time we run something. We are applying the hard lessons learned to finally achieve this reliably.