GitHub - Octoverse 2026: Commits and Pushes Hit New Highs as Developers Build With Agents

GitHub's 2026 Octoverse report, released at GitHub Universe, shows commit and push volume reaching new highs this year, with the company attributing the growth directly to developers building with coding agents rather than writing code by hand line-by-line. This landed the same week GitHub announced it's "redefining the SDLC for the agentic era" with a slate of Universe updates spanning the whole development lifecycle, not just the editor.

Notion - Agents, Rebuilt to Do More for Less

Notion overhauled its AI pricing and infrastructure, introducing a new credit system alongside smarter model routing, long-running agents, and bring-your-own-AI options. The accompanying engineering post, authored by engineering manager Ghita Houir Alami, details how Notion built an upgraded model router called "Auto" that uses pairwise comparisons and calibrated judging to measure model quality at scale, rather than relying on static benchmarks.

LangChain - The Art of Loop Engineering

LangChain published a detailed framework for what it calls "loop engineering," the practice of designing the core agent loop (not just the model or the prompt) as the primary lever for reliability. The post argues that a good model plus a bad harness still produces an unreliable agent, and walks through how stacking and extending loops builds more capable agents, plus how to instrument each layer with LangSmith for debugging.