Artificial Adventures
4 days ago
- #productivity
- #code collaboration
- #AI development
- AI assistants are most valuable for reviewing code and finding bugs, with frontier models performing well.
- Refactoring tasks benefit from AI, reducing the cost of fixing design mistakes, but review can be challenging due to mixed-in incorrect changes.
- Writing code together with AI is problematic because AI makes poor decisions and often disregards instructions, though future harnesses may improve collaboration.
- AI is effective for one-off scripts and small tasks where output can be easily verified, but fails for complex projects like implementing board game rules.
- Search and verification tasks work well when precision is needed, but caution is required for unverifiable answers.
- Brainstorming with AI yields banal suggestions, and existing harnesses are limited, highlighting the need for better control and interaction methods.
- AI subscriptions offer good value, but token-based pricing may change cost-effectiveness, with cheaper models often misaligned or less useful.
- AI use requires adapting practices to manage mental models and verification, with potential for future improvements in model capabilities and tools.