# AI Coding

- [The Spec File as Source of Truth: How to Write Specs That AI Can Actually Implement](https://sdd.sh/2026/05/the-spec-file-as-source-of-truth-how-to-write-specs-that-ai-can-actually-implement.md): Writing specs instead of code is the core premise of SDD — but a bad spec produces bad code just as reliably as a bad prompt does. Here's what separates specs that AI can execute reliably from the ones that waste hours of compute and your afternoon.
- [Google's 75% Threshold: When AI Became the Primary Author of Production Code](https://sdd.sh/2026/04/googles-75-threshold-when-ai-became-the-primary-author-of-production-code.md): Sundar Pichai revealed at Google Cloud Next 2026 that 75% of new code at Google is now AI-generated and reviewed by engineers. That number crossed a threshold most didn't expect this fast — and it reframes every assumption about what software teams look like in 2026.
- [OpenCode at 147K Stars: The Open-Source Terminal Agent That Won't Pick a Side](https://sdd.sh/2026/04/opencode-at-147k-stars-the-open-source-terminal-agent-that-wont-pick-a-side.md): OpenCode has 147K GitHub stars, 6.5M monthly developers, and supports 75+ LLM providers. Here's an honest look at what it gets right, where it falls short, and when it makes more sense than Claude Code.
- [81% vs. 46%: The AI Coding Benchmark That's Been Lying to You](https://sdd.sh/2026/04/81-vs.-46-the-ai-coding-benchmark-thats-been-lying-to-you.md): SWE-bench Verified — the benchmark that put every frontier model above 80% — is contaminated. OpenAI stopped reporting it in February. Here's what actually happened, what SWE-bench Pro replaces it with, and why 46% is a more honest number than 81%.