Reddit review distills coding-agent harness design practices
A Reddit poster synthesized 57 publications on harnesses for software-development and coding agents, arguing that effective harnesses are small, evidence-based and continually pruned. The poster recommended deriving rules from past sessions and failures, keeping instruction files concise, using fewer tools, enforcing rules with executable checks, and removing rules that cannot be shown to help. The post also cited University of Toronto research reporting performance gaps of up to sixfold for the same model under different harness designs, but the underlying studies and figures were not independently verified in the report.
The synthesis offers concrete guidance for improving coding agents through harness design rather than model size alone, including rule pruning, executable guardrails and reduced tool sets.
I think by now most of us have figured out that picking a bigger model isn't necessarily how you get a better agent. What actually matters is how you wrap the model, aka its harness.
And the numbers back this up: Research from the University of Toronto shows the same model produces up to 6x performance gaps depending on harness design, which is wild IMO.
So I got curious what people are actually doing to get those gains. I went through 57 publications to try to get a clearer picture. (FYI I’ve limited the research to harnesses used for software development and coding agents as it’s what I know best.) Here’s what the common practices boil down to, based on that literature research + my own experience:
TLDR: The best harnesses are small, evidence-based and constantly pruned.