← Curriculum track ← Learn hub
Quanta GenAI Curriculum · LLMOps · Basic

LLMOps Basic — 145: iterate incident comms on `Toy workload cost instincts` — memo `306453 [145]`

Lesson 145: Toy workload cost instincts

Focus

Bias toward observable metrics: latency, cost, escalation rate. Token Toy workload cost instincts:145 keeps neighbouring lessons differentiable.

Key ideas

Deep dive notebook

Synthetic drill artefacts

Token CFO scratchpad

- prompt_budget: 1743
- completion_budget: 680
- cache_key: `1215c`

Hypothesis: halving completions moves P95 ~5% — record actuals.

Practice

Practice Attach rollback steps if cost-per-request crosses your guardrail. — 145 Bump 8.