2x, not 10x: Coding with LLMs in 2026
Source: https://obryant.dev/p/2x-not-10x/
Author: Jacob O'Bryant
Date: 2026-07-31
Summary (Chinese)
2026 年 LLM 好用，主要是过了“反馈环里稳定走一步”的门槛。对“结构是否更可维护”“文档该写什么”仍不够准。推测可见未来的增益更多来自围绕现有能力的工具与流程改造，而不是单等下一代模型。可工作的实现曾意味任务 80% 完成，现在只是 20%。
Summary (English)
LLMs' 2026 adoption surge is largely due to passing the 'stable step in feedback loops' threshold. Still insufficient for 'is there a more maintainable structure' or 'does this documentation include the right info.' Predicts future gains from tooling and process improvements around current capabilities, not next-gen models alone. A working implementation used to mean 80% done; now it's 20%.
One-liner
可工作的实现曾意味任务 80% 完成，现在只是 20%
*Imported from Obsidian digest notes on 2026-08-01. Content grounded in fetched source metadata via opencli.*