50% code, 30% AI, 20% manual dataset creation
@mattshumer_
-
Big Models: High Performance Comes at High Cost
By
–
Big model, big examples, big success, big expensive
-
New Claude Agent Framework to Be Open-Sourced Tomorrow
By
–
Built an even wilder Claude Agent than the last two I open-sourced May open-source it tomorrow
-
Claude Prompt Engineer vs Opus to Haiku Tool Comparison
By
–
You might be using claude-prompt-engineer, not opus-to-haiku, btw
-
Few-shot examples boost model performance beyond GPT-4 affordably
By
–
Add a few examples and it’ll outperform GPT-4 by far at a fraction of the cost
-
Claude Haiku recommended for specific use cases
By
–
Try Claude Haiku, likely better for this use-case
-
Claude Prompt Engineer enhances system prompts with better examples
By
–
Yeah, you can use claude-prompt-engineer to get an even better system prompt than this can provide, and use the examples from this system to round things out!
-
Building with Claude 3: Essential techniques you need to try
By
–
If you're building w/ Claude 3, you NEED to try this.
-
Automated Prompt Generation and Testing Workflow for Claude
By
–
It'll then put the examples in the right prompt format for Claude, and test the generated prompt against your initial input example, using Haiku. Lastly, it'll save both the system prompt it generated and the Opus-created examples in a Python file, pre-formatted for generation.