Skip to main content
The QitOS Blog is where we write about how the framework is actually used in research and real agent workflows. The docs explain what QitOS does. The blog is where we explain:
  • why certain design decisions were made
  • what worked in real runs
  • what failed in benchmarks and long-running agents
  • how prompt, parser, tools, memory, and observability interact in practice

​
What we want to publish here

We expect the blog to grow around a few recurring themes:
  • Design notes: why the framework is structured the way it is
  • Field reports: what we learned from long-running coding and audit agents
  • Benchmark reports: GAIA, Tau-Bench, and CyBench observations
  • Prompting and parsing: model harness choices and protocol tradeoffs
  • Observability: how qita turns traces into research artifacts

​
Start here

Why QitOS keeps a single kernel

A short design note on the AgentModule + Engine split and why we avoid hidden second runtimes.

Why reproducible runs matter

A short field note on official runs, normalized benchmark outputs, and best-effort replay.

Tutorial track

If you want the hands-on learning path first, start with the four-lesson tutorial course.