Skip to main content
Technology / Models & Benchmarks
Technology / Models & BenchmarksGemini 2.5, Claude 3.7, and GPT-4.1: I Ran 100 Code-Generation Prompts Against All Three. Here Are the Results.
ArticleI got tired of reading vendor benchmarks. I ran 100 real code-generation tasks from my own work — 50 Python, 50 TypeScri
Read more →
Technology / Models & BenchmarksThe MMLU Trap: Why Your Benchmark-Topping Model Is Failing in Production
ArticleA Fortune 100 insurer selected a model ranked first on MMLU for an adjudication assistant, and within six weeks p95 late
Read more →
Technology / Models & BenchmarksFine-Tuning vs Prompting in 2026: I Tried Both on the Same Real Product Feature
ArticleI took one concrete feature — a Git-style commit message generator — and implemented it three ways: pure prompting, few-
Read more →
Technology / Models & BenchmarksStop Treating Leaderboards as Architecture Guidance: Designing Evaluation for Your Own Stack
ArticleA team blindly chose the top model from public leaderboards and watched latency, cost, and quality collapse in productio
Read more →