I was reading some Anthropic and Google papers about how models achieve successful long-task running and their goals based on a deterministic harness (Claude Dynamic Workflow feature), and decided to create a harness engineering workflow that turned my entrepreneur project, CauseFlow AI SRE Tool, into an open-source solution, using free models.
The papers that I've read:
https://www.anthropic.com/engineering/effective-harnesses-fo... https://www.anthropic.com/engineering/harness-design-long-ru... https://www.kaggle.com/whitepaper-the-new-SDLC-with-vibe-cod...
The CauseFlow AI SRE is Open Source and you can try it here: https://github.com/vinicius91carvalho/causeflow-ai