Descriptions:
Alex Finn walks through 10 Claude Code prompting techniques drawn directly from an official Anthropic prompting guide, presenting each with concrete examples from his own projects. The tips span a range of concerns — from how to write more effective prompts to how to manage Claude’s context window intelligently.
Key takeaways include: always explaining the motivation behind a request so Claude makes proactively aligned changes across the codebase; trusting Claude Code’s auto-compaction feature rather than manually clearing context with /clear or /compact; loading the full project architecture into context at the start of every new session; and using GitHub commits to give Claude persistent memory of past changes across sessions. Finn also covers how to configure claude.md rules to make Claude more or less autonomous — including a ‘default to action’ rule that lets Claude implement ideas without asking permission first, versus a ‘do not act before instructions’ rule for those who want tighter control.
One particularly practical tip: the word ‘think’ is a functional keyword in Claude Code that significantly increases token consumption, so using it casually (e.g., ‘think about why you want to do this’) can unexpectedly triple costs. Finn recommends reserving ‘think,’ ‘think harder,’ and ‘ultra think’ for situations where deeper reasoning is genuinely needed. All rules and prompts mentioned are available in the video description.
📺 Source: Alex Finn · Published January 03, 2026
🏷️ Format: Tutorial Demo
