rules

Paul Graham shares an insight that in startups, exceptions to common rules can be beneficial, and the mark of a good investor is identifying these exceptions.

The article outlines nine prompt discipline rules that reduced wasted thinking in coding agents by up to 70% based on A/B tests on GLM 5.3 and GLM 5.3 Flash models, with a self-testable exam available on GitHub.

The article provides two key rules for using LLMs in writing: avoid using any words suggested by the model and disregard encouragement to maintain an authentic voice.

Share Shadowrocket configuration tutorial, including importing lazy configuration, rule sets, and modules for distinguishing domestic and foreign websites and blocking ads, and fixing message notification issues.

Introduces an open-source project called AGENTS Book Rules, which organizes the core principles of 13 classic programming books (such as Clean Code, Domain-Driven Design, etc.) into rule files that can be directly used with AI coding tools like Claude Code, Codex, and Cursor to improve code quality and architecture design capabilities.

Andrej Karpathy identified 4 behaviors that break Claude Code; a developer expanded them into 21 rules, achieving 82,000 GitHub stars and improving coding accuracy from 65% to 94%.

Andrej Karpathy points out that 90% of Claude's errors come from missing context, and shares 12 rules (such as think before writing code, simplicity first, etc.) that reduce the error rate from 41% to 3%, emphasizing that discipline and effective context management are more important than frameworks.

Andrej Karpathy states that 90% of Claude's mistakes stem from missing context, not model weakness, and provides a set of 12 rules that reduced error rates from 41% to 3% in experiments.

Google updated its spam policies to explicitly prohibit attempts to manipulate its generative AI search results, with penalties for violators.