{"slug": "speed-is-cheap-capability-is-expensive", "title": "Speed is Cheap, Capability is Expensive", "summary": "A developer with 8+ months of heavy agentic coding experience argues that low-cost LLMs like Qwen and Kimi handle 70-80% of software development work, while frontier models only earn their cost on the 20-30% tied to quarterly KPIs. The author notes that once tooling expertise is built, the advantage of expensive models diminishes, and custom harnesses like Claude Code limit flexibility.", "body_md": "## The 80/20 of LLM Costs\n\nHere is how I have come to think about LLMs after 8+ months of heavy agentic coding (200K+ lines, 30 PRs a month). If you are already strong in your tooling, a large chunk of the agentic advantage disappears. The model moves away from compensating for gaps in my knowledge, and just optimizing for speed.\n\nAnd low-cost models deliver that same speed at a fraction of the price. I hit my quarterly KPIs buying the frontier models for only 20-30% of the work. The remaining 70-80% that fills most of my quarter - features, code reviews, bug fixes, hot fixes, general platform maintenance - runs fine on low-cost models (Qwen, Kimi, etc.). The most capable model only earns its worth on that 20-30% tied to quarterly goals, which matches exactly what I have seen tracking my own KPIs.\n\n## The Harness Problem\n\nSo paying $200 to run mostly Haiku-tier work is over-subscribing. I am paying for capability I have already built the expertise around. The harness is limiting too. Claude Code works well when I want everything pre-baked and it is great for someone new to agentic development, but for any real customization, it falls short.", "url": "https://wpnews.pro/news/speed-is-cheap-capability-is-expensive", "canonical_source": "https://blog.grandimam.com/posts/speed-is-cheap-capability-is-expensive/", "published_at": "2026-06-19 00:00:00+00:00", "updated_at": "2026-07-11 08:39:19.351952+00:00", "lang": "en", "topics": ["large-language-models", "ai-agents", "developer-tools"], "entities": ["Qwen", "Kimi", "Claude Code"], "alternates": {"html": "https://wpnews.pro/news/speed-is-cheap-capability-is-expensive", "markdown": "https://wpnews.pro/news/speed-is-cheap-capability-is-expensive.md", "text": "https://wpnews.pro/news/speed-is-cheap-capability-is-expensive.txt", "jsonld": "https://wpnews.pro/news/speed-is-cheap-capability-is-expensive.jsonld"}}