LLM Learning Path
Home Path Warm-up

Habits you can adopt today

One thread per job, precise asks, history and cost.

One conversation, one job

Split “research + draft + translate” into separate threads—often fewer hallucinations and lower token spend (long history costs more and drags context).

Be precise early

State goal, audience, format, constraints, acceptance before iterating randomly.

SOP → Skill

If you do it thrice, capture steps + I/O + failures; start with one hot path, not a giant library.

Skip small talk with the model

Politeness belongs to humans; fluff burns context.

Further reading