Reddit thread points to Humanity's Last Exam

This is less a search for a certificate and more a search for an eval harness, and HLE shows how far benchmarks still are from a true skills report for models.

• –HLE spans 2,500 expert-authored questions across 100+ subjects and includes multimodal items, so it is broad and genuinely hard.
• –Its authors frame it as a measure of structured academic capability, not autonomous research or creative problem-solving.
• –Because the benchmark is fixed and closed-ended, it can rank models but not produce the personalized weak-area report the Redditor wants.
• –That leaves room for subject-specific eval products with scoring, explanations, and gap analysis across coding, security, and DevOps.