To avoid saturated evaluations that only confirm existing capabilities, Notion's team creates difficult test suites they expect to fail 70% of the time. This "headroom" provides a clear signal to model providers about frontier needs and helps the team anticipate where the technology is heading.

A benchmark like SWE-Bench is valuable when models score 20%, but becomes meaningless noise once models achieve 80%+ scores. At that point, improvements reflect guessing arbitrary details (like function names) rather than genuine capability. This demonstrates that benchmarks have a natural lifecycle and must be retired once saturated to avoid misleading progress metrics.

Standard AI benchmarks are an engineering tool for measuring performance. A more scientific approach, borrowed from cognitive psychology, uses targeted experiments. By designing problems where specific patterns of success and failure are diagnostic, researchers can uncover the underlying mechanisms and principles of an AI system, yielding deeper insights than a simple score.

The gap between benchmark scores and real-world performance suggests labs achieve high scores by distilling superior models or training for specific evals. This makes benchmarks a poor proxy for genuine capability, a skepticism that should be applied to all new model releases.

Don't trust academic benchmarks. Labs often "hill climb" or game them for marketing purposes, which doesn't translate to real-world capability. Furthermore, many of these benchmarks contain incorrect answers and messy data, making them an unreliable measure of true AI advancement.

To evaluate an AI model, first define the business risk. Use precision when a false positive is costly (e.g., approving a faulty part). Use recall when a false negative is costly (e.g., missing a cancer diagnosis). The technical metric must align with the specific cost of being wrong.

Don't rely on a simple agreement percentage to validate an LLM judge. If failures are rare (e.g., 10% of cases), a judge that always predicts "pass" will have 90% agreement but be useless. Instead, measure its performance on positive and negative cases separately (e.g., True Positive Rate and True Negative Rate).

To ensure model robustness, OpenAI uses a "worst at N" evaluation metric. They sample a model's output multiple times (e.g., 20) on a given problem and measure the performance of the single worst response. This focuses development on eliminating low-quality outliers and ensuring a high floor for safety and consistency, rather than just optimizing for average performance.

When selecting foundational models, engineering teams often prioritize "taste" and predictable failure patterns over raw performance. A model that fails slightly more often but in a consistent, understandable way is more valuable and easier to build robust systems around than a top-performer with erratic, hard-to-debug errors.