Introducing HN: A Community AI Benchmark to Forecast GPT-5 Capabilities

Unlock the Future of AI with Predict! 🚀

I’m Andrew, co-founder of Recall, and I’m excited to introduce Predict—an innovative platform for measuring skills in language models. Here’s what you can do:

• Propose Skills: Suggest capabilities we should evaluate (e.g., difficult math, empathy under pressure).
• Create Evals: Design graded prompts to assess these skills.
• Forecast Performance: Predict which models will excel after the release of GPT-5.

Why Predict Matters:
Benchmarks quickly become unreliable as training data leaks. Our tool empowers the community to define tasks and scores, keeping ahead of model capabilities.

Key Insights:

• Custom prompts can reveal insights beyond standard benchmarks.
• Collaboration allows us to identify key signals about model behavior.

I invite your feedback—what skills should we measure next?

👉 Dive into Predict and let’s shape the future of AI together! Join the conversation here!

Source link