Fundamentally, AI should be used as a first-point substitution for scanning of documentation, rubber ducky-ing as well as a tool to reduce start-up friction to tasks — but rarely as a tool for bridging gaps in logic ability.
Danger
Any new information provided by AI should be understood and learned before utilising — NEVER implement code or logic you do not understand.
-
Primarily, use AI as a form of rubber duckying. This includes:
- Checking for correct syntax/syntax combinations.
- New syntax should always be understood, and tested with before implementing.
- Clarifying understanding of syntax, methods or other language features.
- “is it correct to say…”
- Debugging issues, only after giving it at least a reasonable amount of time, in exhausting potential angles that can be theorised.
- Checking for correct syntax/syntax combinations.
-
Use it as an aid to explore and discover tools and libraries.
- Though, the discovered item’s native documentation should be scanned for context initially.
-
If all possible branches of possibilities to tackling debugging are exhausted, use AI to minimise time wasted slamming head into keyboard.
- Had many days of frustration lost over simple debugs — I’m opting to consider this as equivalent value to having a mentor, in the interest of accelerating productivity.
Useful Tools:
-
Youtube video summarising tools:
-
GPT3.5 Agent