Uploaded November 2025 | Updated November 2026, 3 hours ago
We optimized Claude Code's system prompt - just its prompt - and achieved +10% boost on SWE Bench. No architecture changes, tool improvements, or fine-tuning.
How? An approach called Prompt Learning. What is Prompt Learning? Itβs an optimization algorithm for prompts. Inspired by RL, it follows an action β evaluation β improvement loop β but instead of gradients, it uses Meta Prompting: feeding a prompt into an LLM and asking it to make it better. We use LLM-generated feedback explaining why outputs were right or wrong, giving the optimizer richer signal to refine future prompts. π Result: measurable gains in accuracy, zero retraining. Use it in Arize AX or the Prompt Learning SDK.
How Prompt Learning differs from DSPy GEPA: arize.com/blog/gepa-vs-prompt-learning-benchmarking-different-prompt-optimization-approaches
More about prompt learning: arize.com/blog/prompt-learning-using-english-feedback-to-optimize-llm-systems
Optimizing Claude.md: arize.com/blog/claude-md-best-practices-learned-from-optimizing-claude-code-with-prompt-learning
Optimizing .clinerules: arize.com/blog/optimizing-coding-agent-rules-claude-md-agents-md-clinerules-cursor-rules-for-improved-accuracy
All rulesets, code, and SDKs are open source:
π Generated Claude Code rulesets: github.com/Arize-ai/prompt-learning/tree/main/coding_agent_rules_optimization/claude_code/act_rulesets/claude-sonnet-4-5-20250929
π Claude Code optimization code: github.com/Arize-ai/prompt-learning/tree/main/coding_agent_rules_optimization/claude_code
π Prompt Learning SDK: github.com/Arize-ai/prompt-learning/tree/main/optimizer_sdk
π Run Prompt Learning in the Arize UI (no-code): arize.com/docs/ax/prompts/prompt-optimization/prompt-learning
We optimized Claude Code's system prompt - just its prompt - and achieved +10% boost on SWE Bench. No architecture changes, tool improvements, or fine-tuning.
How? An approach called Prompt Learning. What is Prompt Learning? Itβs an optimization algorithm for prompts. Inspired by RL, it follows an action β evaluation β improvement loop β but instead of gradients, it uses Meta Prompting: feeding a prompt into an LLM and asking it to make it better. We use LLM-generated feedback explaining why outputs were right or wrong, giving the optimizer richer signal to refine future prompts. π Result: measurable gains in accuracy, zero retraining. Use it in Arize AX or the Prompt Learning SDK.
How Prompt Learning differs from DSPy GEPA: arize.com/blog/gepa-vs-prompt-learning-benchmarking-different-prompt-optimization-approaches
More about prompt learning: arize.com/blog/prompt-learning-using-english-feedback-to-optimize-llm-systems
Optimizing Claude.md: arize.com/blog/claude-md-best-practices-learned-from-optimizing-claude-code-with-prompt-learning
Optimizing .clinerules: arize.com/blog/optimizing-coding-agent-rules-claude-md-agents-md-clinerules-cursor-rules-for-improved-accuracy
All rulesets, code, and SDKs are open source:
π Generated Claude Code rulesets: github.com/Arize-ai/prompt-learning/tree/main/coding_agent_rules_optimization/claude_code/act_rulesets/claude-sonnet-4-5-20250929
π Claude Code optimization code: github.com/Arize-ai/prompt-learning/tree/main/coding_agent_rules_optimization/claude_code
π Prompt Learning SDK: github.com/Arize-ai/prompt-learning/tree/main/optimizer_sdk
π Run Prompt Learning in the Arize UI (no-code): arize.com/docs/ax/prompts/prompt-optimization/prompt-learning










