Coding-agent harnesses cost more without better benchmark scores
Harness swaps changed outcomes about as often as reruns, while system prompts made some tasks cost up to three times more.
46
Product
mini-SWE-agent
- Mentions · 7d1
- Mentions · 30d3
- All mentions4
- First seenSep 1, 2026
- Last mentioned6d ago
Timeline
4 articles · newest firstShow HN: Parsimony – an open-source benchmark for how much code AI models add
Parsimony ranks AI coding models by net syntax units added on SWE-bench Verified patches.
34
SWE-Duel: Towards a Self-Scaling Arena for Coding Agents
SWE-Duel has coding agents author and repair hidden bugs so difficulty scales with participants.
62
Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems
Paper models multi-agent LLM orchestration as a bilevel game, proving transcript-only gating limits and introducing grounded-memory SRMA.
32
Appears with
- Zhipu AIorganization · 1
- GLM-5.3-Flashai model · 1
- Deepseek-V4-Flashai model · 1
- OpenCodeproduct · 1
- HY4-Previewai model · 1
- Parsimonyproduct · 1
- Claude Opus 4.8ai model · 1
- Codex CLIproduct · 1
- Kimiai model · 1
- SWE-bench Verifiedproduct · 1
- Claude Opus 5ai model · 1
- Claude Codeproduct · 1
- SWE-benchproduct · 1
- Qwen3.8ai model · 1
- GPT-5.5ai model · 1
- Qwen3.6ai model · 1
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.